Dream Menu
Scotland eats
Every venue named by a guest on the Dream Menu interview series — and the dishes they named it for.
The dish rankings below are the response data's own category counts, reproduced in full and in order — every guest names a dream starter, main, side, dessert and drink. Tap a category under "Browse the map" to pull the venues that logged it.
Most popular dishes — all 1,230 interviews, ranked
Counts and share of all responses, matching the response spreadsheet exactly. Answers too vague to classify are the only rows omitted.
Where guests send people next — venue types
Each guest names the next stop on the dining scene. Recommendation counts by venue type, from the response spreadsheet.
Browse the map by category
Dishes tied to a named, located venue, rolled into supply categories. Tap one to list its venues.
Items named at more than one venue
Where the same thing surfaces independently across venues.
Where each category concentrates
The territory logging the most records in each of the top categories.
Alongside the main interview series, Dream Menu runs themed one-off surveys. These are small, fast samples — tens of responses, not the 1,230 behind the main dataset — so read them as flavour and direction, not hard market share.
Dream Fringe · 38 responses · Aug 2026
Edinburgh Festival Fringe edition, run with performers, producers and Fringe regulars. Where they drink, eat and see shows.
Fringe bars named
Venue preference: the intimate room wins overwhelmingly — small, low-ceilinged, often underground. Theatre, outdoor and arena settings trail well behind. Repeatedly described as dark, cramped and slightly unsound; that is the compliment.
Fringe fuel: coffee first and in volume, then water and electrolytes, with a protein bar or pastry as the day goes on. Energy drinks and meal deals feature heavily.
Late-night orders: chips with cheese and curry sauce is the single most repeated answer, followed by kebabs, shawarma wraps, pizza slices and loaded fries.
Edinburgh food venues named by Fringe respondents: Mary's Milk Bar, Mosque Kitchen, Palmyra Pizza, Nile Valley, Civerino's, Origin Coffee, Buffalo Truck, Wings, Ting Thai, Oink, Tempting Tattie, The Clamshell, Frites, Cargo Spud, Umai, Petra, Home Bakery, Stories Bakery, The Bow Bar, Kebab Mahal, Captain's Bar, The Tipsy Midgie, Noodles & Dumplings, Cafe Piccante, Amorino, Bistro Coco, Contini Cannonball, Jarvis Pickle, Yum Yum Kebab House, Alby's, Pizza Geeks, Bucks Bar, £2 Pizza Slice, Hajar Shawarma (Leven St), Bowls on Forrest Road.
Dream Cocktail · Jun 2026
Respondents build a cocktail from base spirit to garnish. Directly relevant to drinks range planning.
The signal: agave and rum lead the base spirits, well ahead of gin — a reversal of the gin-led decade. Scotch appears, but as a cocktail base rather than a dram.
Rims and wild cards are the fixation: Tajín, chilli, smoked and flavoured salts recur constantly, alongside genuinely left-field savoury elements — miso, gochujang, Marmite, beef-fat caramel, smoked chicken-skin salt, brine. Sweet-only cocktails are rare; almost every entry pairs sweetness with salt, smoke, acid or heat.
Format: highballs and rocks glasses dominate over coupes. Long and refreshing or short and punchy — very little in between.
Dream Sandwich · Apr–May 2026
Bread up. Includes an iced-latte build, making it a useful cafe and daytime-trade companion to the cocktail set.
The signal: enriched and crusty breads beat plain sliced entirely — brioche, focaccia, sourdough, ciabatta and the Scottish morning roll. Provolone and mozzarella outpace cheddar.
The crunch rule: nearly every entry adds a fried or crisp element inside the sandwich — potato scones, hash browns, crisps, plantain chips, croquettes, crispy shallots. Texture contrast is the near-universal design principle.
Heat and pickle: hot honey, gochujang, chimichurri, scotch bonnet and jerk sauces recur, almost always balanced by something pickled.
Every figure on this platform can be traced to its source. This page sets out where each number comes from, how confident you should be in it, and what it does not show. If a number here is ever challenged, the answer should be in this page.
The dataset
| Main interview series | 1,230 responses | Sept 2025 – Aug 2026 |
| Operator survey | 55 operators | 2026 |
| Dream Fringe | 38 responses | Aug 2026 |
| Dream Cocktail | small sample | Jun 2026 |
| Dream Sandwich | small sample | Apr–May 2026 |
| Venues mapped | 615 | live |
| Dish records | 454 | live |
Participants consent to editorial use and to anonymised, aggregated analysis including reports shared with business audiences. No respondent is identified, quoted attributably or profiled. Contact details are collected for editorial purposes and appear nowhere in this platform.
How far to trust each number
Figures are graded by how much interpretation sits between the response and the number.
A Straight from the responses
Counted with no interpretation. Quotable without qualification.
- Course rankings across all 1,230 interviews — starters, mains, sides, desserts, drinks — reproduced in the Trends tab in full and in order.
- Venue-type recommendations: what kind of place guests send people to next.
- Response counts and dates for every survey.
Each course totals slightly under 1,230 because unclear or blank answers are excluded: starters 1,204 · mains 1,195 · sides 1,179 · desserts 1,186 · drinks 1,194. Excluded rows are answers too vague to classify, not rows dropped for being inconvenient.
B Structured but derived
Sound, but the structure is an editorial judgement. The method is defensible; the precision is not the point.
- Supply categories (19) — the 454 venue-linked dish records rolled into categories a distributor recognises, by a published keyword classifier. Boundaries are judgement calls, applied identically to every response and open to inspection.
- Territories (16) — venues grouped commercially rather than by postcode. Currently placeholders, to be replaced with real sales territories.
- Course per dish — Main 171 · Dessert 94 · Starter 82 · Drink 71 · Side 36. Inferred from the dish, not stated by the guest.
C Inferred — flag before using
Useful, but say "inferred" out loud before anyone else says "invented".
- Venue type — read from the venue name first, logged dishes second. 133 of 615 venues fall through to "Restaurant (general)" because nothing was decisive. Individual assignments are arguable.
- Nine venue locations were geocoded from general knowledge rather than the main geocoding pipeline: Celentano's, Lychee Oriental, Cafe Strange Brew, Dennistoun Bar-B-Que, Errol's Hot Pizza, Alma Mexicana, Bolt Burger, Ardnamurchan, Brutti Compadres. Verify before operational use.
D Not ready — do not present
Venue-level mention ranking. The obvious headline — Scotland's most-recommended venues — is deliberately absent. The raw venue field currently holds the same venue under several spellings, so any ranking built on it today would not survive scrutiny. A deduplication tool exists and works; it has not yet been run against the full export with human review.
Movement over time and demand versus menu presence are supported by the data but not yet produced.
Questions this data invites
Is this representative of Scottish diners?
No, and it isn't trying to be. It is a self-selecting, food-literate audience — people who follow and post about food closely. That skew is deliberate: this group's preferences tend to surface in the wider market later. It is a leading-indicator panel, not a representative sample.
Is this sales data?
No. It is stated preference, gathered in a fantasy-menu framing. It does not measure what sold, what was ordered, or what appeared on menus. It is strongest read alongside sales data, not instead of it.
A category scored low — does it not sell?
No. It means guests didn't volunteer it. Absence of mention is not absence of demand, and this matters most for unglamorous high-volume lines. Nobody's dream side dish is a bread roll.
Aren't the specials samples very small?
Yes. Fringe is 38 responses; cocktail and sandwich are smaller. That is why those sections show counts rather than percentages — percentages on a base that size imply a precision the data does not have. Read them as direction only.
Is 55 operators enough?
It is a small base. One operator is worth 1.8 percentage points. Operator findings are stated as direction, never as precise shares.
How independent is this from its sponsor?
The sponsor has had no influence over methodology or findings. Questions covering beer and wine are a natural evolution for a food and drink interview series; they are open text, with no brand prompting and no supplier lists to choose from. The answers are whatever respondents say. That independence is precisely what makes the data commercially worth anything.
Can the underlying data be shared?
Aggregate outputs, yes. Raw responses, no — participants consented to aggregated analysis, not to individual answers being passed to third parties.
What this data cannot say
That a share of Scots want something — it is the community, not the nation. That demand is proven — it is indicated. That any venue is the most popular, until the mention tally is run and reviewed. That sales of anything are rising — there is no sales data here. Any percentage on a base under about thirty without stating the base alongside it.
The dataset at a glance
Independent interview responses from Scotland's food community — chefs, venue owners and food creators — collected by the Dream Menu series. A further 34 venues named abroad are counted but not mapped.
What the data says
Territory coverage
Mapped venues and dish records by commercial territory. Where coverage is thin, so is the interview pipeline — a target list, not just a gap.
How to read this platform
Restricted section
This section contains the full insight suite — trend analysis, venue rankings and commercial reporting — and is shared with partner organisations rather than published openly.
Interested in access? See what's included
| # | Venue | Type | Address | Logged dishes |
|---|
Dream Menu — venue advocacy from Scotland's food community. Stated preference, not sales data.
Named only