VacationSelector

How it works

Data first, words second

Every destination is described by hundreds of figures pulled from openly licensed sources: 30-year climate normals, satellite sea temperatures, official tourism statistics, price levels, restaurant and nightlife counts from Overture Maps, notable places from Wikidata weighted by how many Wikipedia language editions write about them, and counts of museums, beaches, parks and playgrounds from OpenStreetMap. The figures are stored with their source and licence, and each page names the source next to the number.

Every score on this site is calculated from those figures by a formula, and each one is shown with the figures it was calculated from. No score is a judgement, and none of them depends on which other destinations we happen to have: a city’s score is worked out from its own measurements and from anchors fixed in advance, so it does not move when we add somewhere new.

A language model is used for exactly one thing: phrasing short summaries of differences the data already shows. It is never allowed to supply a figure, and any block it cannot ground in data is simply not shown.

What the axes can and cannot see

Each axis is only as good as the figures behind it, and several are worth knowing about before you read a score. Figures on this page are quoted as they are measured, in kilometres, hectares and degrees Celsius, whichever units you have chosen elsewhere. That is deliberate for the radii, which are not descriptions of a distance but the definition of the measurement: the same disc is drawn around every destination in the world, and restating it in miles would suggest a different one was used. It applies to the rest for the plainer reason that this page explains how the numbers were made rather than reporting them. Every figure on a comparison page follows your choice. “Places to eat” counts restaurants within 2km of the centre and the kitchens they list, so it measures how much choice you have and how broad it is. It is not a measure of how good the cooking is: no open dataset carries that, and a city famous for doing one thing superbly will score like any other city its size.

“Museums & heritage” works the same way. It counts the museums mapped within 10km of the centre, how many of those sit within 2km, and the whole UNESCO inscriptions within 50km. That is the size and the concentration of the offer, not its importance, and the two are not the same: the wider count grows with how much of a city falls inside the radius, so a compact centre or one ringed by water is measured over less ground than a sprawling one.

“Value for money” rests on a daily budget published per country rather than per city, so two cities in the same country carry the same figure and the comparison between them says nothing. Where that applies, the comparison page says so.

“Beaches & swimming” asks whether you can swim, not whether the water is salt. A city scores here when the satellite temperature record resolves a body of open water beside it, which is why Chicago is measured on Lake Michigan at 22.5°C in August while Berlin and Vienna score nothing on lakes and river arms too small for the record to see. That cut-off is the satellite grid rather than a judgement about the water, and it is not tidy: it reaches the Great Lakes and the Caspian but not Lake Geneva or Balaton. A city on one of those would score zero here for want of a measurement, not for want of swimming.

“Nightlife” counts bars, pubs, cafes and beer gardens as one thing, because the words mean different premises in different countries: an Italian bar is usually mapped as a cafe and a Spanish one as a bar. In four cities the count comes out higher than the number of restaurants, and we checked whether those were really eating places filed under the wrong heading. They are not: in Genoa 4 per cent of them carry any sign of serving a meal, against 43 per cent in Berlin. Where there are more places to drink than to eat, that is what the streets hold. Three of the four are southern European, Genoa, Granada and Seville; the fourth is Tokyo, which is a useful reminder that the pattern is about how a city drinks rather than about where it sits.

“Places to eat” counts sit-down restaurants only. Takeaway and counter-service places are a separate category the score does not read, and how much of a city’s eating happens there varies enormously: they add 8 per cent on top of Lisbon’s restaurant count and 57 per cent on top of Genoa’s.

“Nature & day trips” measures two things: how much of the land within 50km is protected, and how many hectares of park lie within 2km of the centre. Both are measured as ground rather than counted as objects, and the difference is larger than it sounds. We used to count parks, until we checked what that reads: Faro maps 42 parks across 18 hectares and Muscat 6 across 12, roughly the same amount of park, and the count made one look seven times greener than the other. What it was really measuring is how finely a country’s volunteers divide green space up. Hectares do not care. The axis has one known blind spot we could not close: it counts large inland lakes as protected ground where a reserve is drawn over the water. Amsterdam is the destination this affects, and materially: the Markermeer and the IJsselmeer alone are 46 per cent of its protected area, and counting the smaller waters with them takes it past half. We built six ways to strip the water out and measured every one; each removed real reserves along with the lakes, including a wood and a forest reserve, because the map uses the same tag for a lake and for the marsh of a wetland reserve. A smaller error we can describe beat a larger one we could not see.

Everything counted from OpenStreetMap depends on how thoroughly volunteers have mapped a place, and that varies far more between countries than most of the differences the counts are used to describe. Within 2km of the centre Shanghai carries 1,013 mapped objects against Rome’s 2,529 and Hong Kong’s 1,206, so its restaurant and playground counts understate the city rather than describe it. Hong Kong is the telling comparison: it is mapped from the same national extract as Shanghai and is nearly as dense as Rome, so the gap is not a fact about the country. We check this by measuring the same feature two ways in two countries before trusting a gap, and where the gap turns out to be the map rather than the place, it is written down rather than quietly scored around. Two such gaps were large enough that we changed the source instead: restaurant and nightlife counts now come from Overture Maps, whose commercial coverage does not depend on volunteer density, and the sights count comes from Wikipedia notability, which no mapping convention owns. Wikipedia has its own lean, towards Europe and towards cities made of many separately famous buildings, so a city whose heart is one great ensemble, a medina or a bazaar, reads lower there than a visitor would score it, and the caveats on its pages say so.

Your weighting, your verdict

The match score is a weighted average of nine axis scores. The sliders set the weights, the calculation runs in your browser, and the link records your weighting so a shared comparison shows exactly what you saw. The underlying scores never change with the sliders; only their importance does.

Sources and licences

Rubric v2 · data snapshot 2026-08-31. Estimates (daily budgets, flight times) are labelled as estimates wherever they appear.