#CityAttention · Sentiment · Experience · Consumption · AccessScore
The open-data ranking of the world's 100 great travel cities. Every input is public, every score is reproducible, and the full dataset ships with the ranking. No panels, no black boxes.
Five dimensions, weighted 25 / 15 / 25 / 20 / 15, frozen before any score was computed. Heavy-tailed signals are log-transformed, everything is min-max normalized to a 1-10 band, and the weights are frozen per edition.
Wikipedia pageviews for each city across the 14 top traveler languages plus English Wikivoyage, 12 full months, bot traffic excluded. Reddit travel-forum mention volume adds the conversation layer.
Top Reddit travel-community posts about each city, classified positive, neutral or negative as a travel destination by an LLM, then buzz-weighted so a handful of glowing posts can't outrank a widely loved city. Sample sizes are published per city.
OpenStreetMap points of interest within 10 km of the city center, in five buckets: culture, historic, food, nightlife and green space. The same radius and the same query for every city.
Where travel money actually works: OpenStreetMap shopping supply (department stores and malls, fashion, beauty, convenience) and stay supply within 10 km, price level from World Bank PPP data, and whether the country runs a verified tourist tax-refund scheme, source-checked per country.
Visa openness of the destination (share of world passports admitted visa-free, on arrival or with an eTA) and non-extreme-weather days per year from the Open-Meteo 2025 archive, a rule recalibrated in v1.1 so tropical cities are not structurally punished.
Known limits, stated plainly: the sentiment sample is small (top Reddit posts per city, N published) and skews English-speaking; OpenStreetMap is mapped denser in Europe than parts of Asia and Africa; Wikipedia attention can spike on news, not just travel intent; visa and price signals are national figures applied to cities. We publish these limits because an index you can audit beats an index that pretends to be perfect.
Missing signals are disclosed and re-weighted, never imputed. Google and Tripadvisor ratings are deliberately absent: their terms prohibit storing or aggregating ratings into derived indices. The city set itself is the one editorial act, published and stable across editions. Full formulas live in the methodology file that ships with the data.
The entire dataset behind the ranking, including every raw signal, is free to download and reuse with attribution. If you re-compute the index and get different numbers, tell us: that's the point.