Methodology.

The two halves

Each half is built from several signals. We name the signals here. We do not publish the exact weights — partly because they are still being tuned, and partly because the editorial point of the score is harder to copy than the recipe.

Beauty

• Heritage architecture density — counts of OpenStreetMap heritage features (castles, churches, towers, listed buildings, archaeological sites) within 2 km, ranked within each country to control for OSM tagging biases that structurally favour data-rich countries.
• Everyday amenity density — concentration of museums, restaurants, cafés and places of worship within 2 km (walking distance). The signal that distinguishes a lived-in small city from a pretty-but-empty hamlet. Capped outside Europe, where OSM amenity tagging is structurally thinner beyond dense urban cores, to avoid penalising genuinely rural picks.
• Natural setting — proximity to water bodies (rivers, lakes, sea) and to mountains, computed against densified OSM polylines and SRTM-derived peaks.
• UNESCO & protected areas — counts within 50 km of UNESCO World Heritage sites and IUCN protected areas.
• Aesthetic score — a CLIP+Ridge model trained on a hand-labelled corpus, scored against the destination's primary photograph from Wikimedia / Wikivoyage / Flickr.
• Photographic diversity — the number of distinct photographers contributing photos within 2 km on Flickr, divided by population. The thesis: lots of different people each taking a few photos beats a few people each taking lots, because it captures genuine appreciation rather than a single influencer's bias.

Economic need

• Income deviation — local GDP per capita (or equivalent income proxy) versus the country mean, computed at the smallest available subnational unit. The traveller's marginal dollar is worth more here.
• Unemployment deviation — local unemployment rate versus the national mean.
• Hotel occupancy proxy — under-capacity areas score higher than oversaturated ones, where we have the signal.

The penalty

Beauty and economic need are not the whole story. A place can be beautiful, useful, and already crushed. So we subtract a tourism-intensity penalty built from:

• Wikipedia pageviews — both absolute and per-capita (per resident).
• English-share of pageviews — the fraction of Wikipedia views in English vs. the destination's primary language. A high English share is a leading indicator of tourism saturation.
• Hotel density — number of hotels per km².
• Sitelink count — how many language editions of Wikipedia / Wikivoyage have an article about the place.

The penalty is subtracted from the beauty + need composite. This is why a famously photogenic but heavily-touristed place ranks below an equally-beautiful but quieter neighbour.

Normalisation

Every signal is normalised within country, not globally. A small Italian hilltop town competes with other Italian places, not with rural Mongolia. This sidesteps a long list of structural biases — OSM tagging density (Germany over-tags heritage), GDP units (log vs. raw), and uneven coverage for non-OECD subnational statistics.

Country-tier floors ensure that smaller countries still get a representative sample on the site even when their absolute WWS scores are lower than the biggest contributors.

The launch dataset covers 12,164 destinations across 233 countries and territories, grouped into 18 themes.

What this can't tell you

• Whether you will love it. The score is a point of view, not a verdict.
• Whether it's safe right now. Travel advisories trump any algorithm — check the US State Department's travel advisories list (or your own country's equivalent) before booking.
• Whether the data is fresh. These were the most recent data available for our version 2.0 launch in May 2026.
• Whether the photo is the best photo. We use a single primary photograph per destination from sources including Wikimedia, Wikivoyage, and Flickr; better photos exist for many of them.

How tier badges work

Every destination on Wandering Well already made the cut. We start from millions of populated places in OpenStreetMap and GeoNames, apply the score, and surface roughly 12,000 — the slots each country's filter let through. Everything here is a pick. The medal is a further distinction within that already-curated set, computed by within-country rank so a Bhutan pick doesn't have to compete with Italy on absolute terms:

• Wander Well Gold — the top quarter of our picks in each country.
• Wander Well Silver — the next ~38%.
• Wander Well Bronze — remaining 38% of beautiful places that rated as Wandering Well destinations for each country.

Tiers are how the score surfaces on the site. The raw composite is real — beauty plus economic need, minus the tourism penalty, all normalised within country — but we don't print the number anywhere, because a number to one decimal place implies a scientific precision the underlying signals don't justify. The tier badge says what the score is for: a perspective on whether a place is worth your time, not a verdict to compare against somebody else's.

Sources

World Bank · UNESCO · Geoapify · OpenStreetMap · Eurostat · Eurostat hotel-occupancy statistics · OECD Regional · US Census ACS · UK ONS · Statistics Canada · Global Data Lab · ILOSTAT · IBGE · Wikipedia · Wikivoyage · Wikidata · Flickr. Per-destination photo attributions on each destination page.