How the scores work.

Who is listed

A spot is included only if a named publication lists it: mostly Condé Nast Traveller India city round-ups, plus Outlook Traveller for Kolkata. Each article is linked with its publication date and the date we checked it. We do not copy review text.

Ratings

We record a rating only when it is publicly visible without signing in. In this edition that means the rating and rating count in the schema.org data of a public Zomato listing page. Google Maps and Swiggy values are marked not verified: Google renders ratings with JavaScript, Swiggy refused plain requests, and we do not bypass access controls.

Composite score

For each verified platform, normalised rating r = rating × 5 ÷ platform maximum. Bayesian average = (C × m + n × r) ÷ (C + n), where n is the number of ratings, C = 500, and m = 4.2222, the rating-count-weighted mean of every verified rating in this dataset. Platforms are then combined weighted by n. This pulls thinly rated places towards the average, so a handful of ratings cannot outrank thousands.

Worked example: a 4.4 rating from 11,473 ratings gives (500 × 4.2222 + 11,473 × 4.4) ÷ 11,973 = 4.39.

Authenticity profile

Style, cooking method and heritage appear only when a cited source states them, labelled “as reported”. A blank means not verified, not absent.

Limits

Platform ratings mix dine-in and delivery, outlets and years. Press lists reflect their writers’ choices on their dates. Nothing here is a tasting, a visit or an endorsement, and no place paid to be listed.

Aggregated, not first-hand. Aggregated from public listing data and press mentions. Fuelled by Biryani has not visited, tasted or photographed these places. Data checked 2026-10-15. Ratings change; open each source for the current value. Corrections: [email protected].