How every number is built.

The corpus

14 cities live, 1 in extraction. 1,524 venues, sourced from Resy and The Infatuation's city guides. 73,628 dishes, read from full menus and parsed into ingredient tokens. Menus were snapshotted Jul 13, 2026; search interest was fetched Jul 16, 2026. Everything on the home page is that snapshot — not a live feed.

For exactly which menu tokens the flavor radar removes and why, see what the radar filters out →

Credibility scoring (max 90)

A venue's score is credentials, not vibes. It is the venue's single strongest culinary award — Michelin ★★★ 60 · ★★ 51 · ★ 42 · World's 50 Best top-50 38 · James Beard winner 34 · 50 Best 51–100 30 · James Beard nominee 20 · Michelin Bib Gourmand 16 — plus an editorial stack capped at 10 (NYT Top 100 5 · 50 Best Discovery 4 · Eater / The Infatuation / Bon Appétit 3 each · local city lists 2 · OpenTable / Yelp 1).

• Awards & listsmax 60
• Chef pedigreemax 10
• Conceptmax 10
• Buzzmax 10

Awards 60 + chef pedigree 10 + concept 10 + buzz 10 = max 90. One Michelin star (42) outscores the most a venue can reach with no award at all — credentials always outrank buzz. Tiers: Elite ≥ 68 · Strong 48–67 · Solid 30–47 · Emerging below 30. 988 of 1,524 venues clear the scoring bar; every venue page shows its own four-bar breakdown.

The flavor rankings

Trend Score — the composite, out of 100. Five signals, each paying out of its own max by where a flavor ranks among all 867 flavors we track — top of the pack takes nearly full points, mid-pack about half. In short: 60 of the 100 points are what kitchens do, 40 what the public searches. On the board, click any flavor for its own breakdown.

• Menu adoption — up to 25 pts — venues serving it right now.
• Elite kitchens — up to 20 pts — the share of its venues that are elite-tier.
• City spread — up to 15 pts — how many of our cities carry it.
• Search level — up to 15 pts — 30-day average US search interest.
• Search growth — up to 25 pts — last week's search interest minus the first week's.

Google Trends — public momentum. 30-day US search interest, via SerpApi. Ranked by growth: last week's average interest minus the first week's. Flavors with almost no search volume (level under 3) can't rise on noise, so they sort last.

Classic — menu adoption. Flavors ranked by how many venues across our cities serve them right now. No search data, no momentum — pure presence on the plate. Ties break on dish count. Not currently a tab on the world board — it still breaks ties in every ranking, and it is the ordering city radar pages use.

Reading the trend lines

Each curve is scaled 0–100 against its own 30-day peak — so they compare shape and momentum, not absolute search volume. Two flavors both topping out at 100 are not equally searched; each is simply at its own busiest point.

Search interest is US-only (Google Trends via SerpApi). Even for a Tokyo or Paris venue, the curve measures American demand for a flavor, not local demand.

The sources

Every credential traces to a named authority. The venue and menu data behind the corpus comes from Resy and The Infatuation; scores are cross-referenced against these:

Known limits

• Search interest is US-only — it does not capture demand in a venue's home market.
• Menus are a point-in-time snapshot (Jul 13, 2026), not a live feed; a dish added yesterday isn't here yet.
• 988 of 1,524 venues are credibility-scored — the rest simply lack a scoreable credential in our sources. It doesn't mean they're bad.
• Chef pedigree, concept and buzz are assessed, not awarded — which is why they cap at 10 each against an award's 60. The score measures credentials, not how good dinner is.