Vid

Method

How Vid finds and ranks places.

No paid placements, no invented review stars, no black box. Here is the whole pipeline: the tags we ask OpenStreetMap for, the signals that move a place up, and the licence check behind every photo.

Last updated .

Short version. A place earns points for being covered by Wikipedia in many languages, for being read, for being heritage-listed, for having a Wikidata item and a free photo, and for being close. Type gives a modest prior. A railway station earns nothing — it is not a destination.

What gets searched

Vid uses a curated catalogue of OSM tags rather than “every named object”, which would return benches, driveways and bus stops. The main groups are:

  • History: castles, forts, ruins, archaeological sites, battlefields, manors, palaces, monasteries, abbeys, priories, mines, mills, historic ships, towers and city walls.
  • Sacred: churches, cathedrals, chapels, mosques, synagogues, temples and monasteries.
  • Landmarks: attractions, public art, bridges, fountains, libraries, theatres, opera houses, town halls, markets, stadiums, water towers, gasometers and piers.
  • Views: viewpoints, peaks, cliffs, towers, lighthouses, observatories and water towers.
  • Nature: parks, gardens, nature reserves, waterfalls, caves, beaches, springs, volcanoes, glaciers, valleys, named lakes, woodland and national parks.
  • Family and food: zoos, aquariums, theme parks, water parks, mini golf, restaurants, cafés, bars, pubs, bakeries and ice-cream shops.

Deliberately excluded: railway stations and halts. A station is a waypoint, not a destination, and major termini used to crowd out real sights in dense cities such as Vienna. Hotels, parking, playgrounds, sports pitches and generic historic=yes objects without a notability tag are excluded too.

OSM tagging is incomplete. To catch famous places that mappers tagged thinly, Vid also queries Wikidata for notable items of the same types and merges them in when the OSM result set is sparse.

How the score is built

Every result is scored after enrichment, once, and capped at 200. The exact weights:

Ranking signals and their weights
Signal Points
Fame — Wikipedia language editions 6 / 12 / 20 / 30 / 42 / 56 / 70 for 1 / 3 / 6 / 11 / 21 / 41 / 81+ editions
Popularity — average monthly Wikipedia reads 4 / 10 / 18 / 28 / 36 / 44 for 500 / 2k / 5k / 10k / 25k / 75k+
Type prior (castle, museum, church, park, café…) 46–96 base × 0.35 = 16–34
Heritage designation (OSM or Wikidata) +12; World Heritage +8, or a high heritage grade +5
Wikipedia article / Wikidata item +8 / +3
Photo with a commercial licence +6
Description +4
Website / opening hours +2 each
Step-free access / outdoor seating / cuisine +1 each
Proximity to the chosen centre 0–8, scaled to the radius
Chain brand −10

A Wikidata item that describes a city, district, street or person is detected from its P31 type and never exports its fame to a mis-tagged memorial or building. You can always switch the list to plain nearest distance.

Photos, licences and credit

A photo is shown only when Wikimedia's metadata says the file is free for commercial reuse: CC0, public domain, CC BY, CC BY-SA, or Commons' plain Attribution licence. Non-free, NC/ND, GFDL-only and files with usage restrictions are dropped. The same check runs for Wikipedia lead images and for the Wikidata P18 fallback.

Images are served as compressed Wikimedia thumbnails (640–800 px), never the full original. Every photo carries its author, licence and a link to the Commons file page in the popup and the detail view.

Where the data comes from

  • OpenStreetMap via Overpass — places, tags, coordinates and heritage (ODbL).
  • Wikipedia — descriptions, article links and lead images (CC BY-SA 4.0).
  • Wikidata — notability, type, heritage, sitelinks and the P18 image (CC0).
  • Wikimedia Commons — per-file photo licences and author credit.
  • OpenFreeMap — vector map tiles and styles.

The full attribution and commercial-use audit is in THIRD_PARTY.md.

See the method on a real map.

Create a free account, pick a place, and compare the ranked list with the nearest list. The same scores, the same data, no ads in between.