OpenStreetMap Hotels via Overpass Turbo

The world's volunteer-built map keeps a lodging inventory that commercial directories would charge to imitate: 452,341 objects tagged tourism=hotel, counted at the August 2026 tag snapshot as 255,658 point nodes, 190,479 building-footprint ways and 6,204 relations, beside 329 explicit tourism=resort entries and hundreds of thousands of hostel, guest-house and apartment records. OpenStreetMap Hotels via Overpass Turbo is that inventory made addressable - selected by tag, bounding box or named area and answered with individual properties carrying all their tags.

Each row travels with whatever mappers recorded: name near-universally, then operator and brand for chain attribution, stars, rooms and beds where documented, split addr:* address components, phone, website, internet_access and wheelchair. It is deliberately not a bookings or rates feed - no occupancy, no pricing, no reviews. This is where lodging stands and who runs it, at whatever slice you ask for. Within Datadory's catalog of 1,744 datasets across 159 viable industries, this slice holds ten datasets and this record rates 8/10 against a 7.81 catalog-wide average. Get a sample of this dataset and read the rows before you commit.

Three real central-London hotel nodes from the August 2026 research pass, chosen because together they show the completeness gradient better than any single row could:

# three central-London hotel rows, exactly as delivered - August 2026 research pass

# row 1 - the full-enrichment end of the spread
element    : node 413596123          position : 51.5174483, -0.1304949
tourism    : hotel
name       : STG Hotel Oxford Street
operator   : St Giles Hotels
address    : 12 Bedford Avenue, WC1B 3GH, London
phone      : +44 20 7300 3000        website  : stghotel.com
email      : lbook@stghotel.com      hours    : 24/7
internet   : wlan                    wheelchair: limited

# row 2 - a housename doing the work of a brand tag
element    : node 346430283          position : 51.5083992, -0.1247180
tourism    : hotel                   name     : The Clermont
housename  : Charing Cross Hotel
address    : WC2N 5HR, Strand, London
wheelchair : yes                     check_date: 2025-08-20

# row 3 - the sparse end: name, street and city, equally valid
element    : node 665982792          position : 51.5196094, -0.1239586
tourism    : hotel                   name     : The Buckingham
address    : 39 Bedford Place, London, GB

# 452,340 further rows repeat this envelope - enrichment varies, shape does not

Read them together. Row one is full enrichment: operator named, phone, website, email, opening hours, wifi and a wheelchair flag of limited - a property a directory could ship tomorrow. Row two shows a subtlety most extractions flatten away: The Clermont trading from the building whose addr:housename still reads Charing Cross Hotel, accessibility marked yes, freshness dated. Row three is the sparse end - name, street, city - and it is an equally valid row, not a defect.

Every remaining row repeats this exact envelope: element class, stable id, position or resolvable footprint, tag bundle. Enrichment varies by design; the shape never does, which is why joins land on the first attempt.

Sixteen columns define the delivery envelope, every definition verified during the August 2026 cataloging pass against live records and the published tagging documentation rather than inferred from headers.

type and id fix what the row is and keep it addressable across syncs - the quiet workhorse that ties consecutive deliveries and cross-checks against other layers derived from the same map. lat and lon place point features while ways and relations carry their own member geometry for footprint resolution. tags.tourism decides whether a row is a hotel, a hostel or a resort out of the accommodation family, and tags.name supplies the label everything downstream displays. The commercial pair follows: tags.operator names who runs it, tags.brand makes chains countable - then the capacity trio of tags.stars, tags.rooms and tags.beds, the split tags.addr:* address components, tags.website, tags.phone, and the doorstep attributes tags.internet_access and tags.wheelchair.

Three surfaces fold under additional fields on request: companion-tag bundles such as email, opening hours and check-in times; polygon geometry resolution for the footprint elements; extended accommodation values, named-area scoping, change slices and format parity. They ship confirmed against the geographies and property families you name during sample preparation instead of promised blind.

• Geography - worldwide wherever the map has been built: every continent represented, densest in Europe and North America, thinner where contributor populations are smaller. Density is stated per region rather than averaged over, with completeness flags surfacing thin spots in evaluation rather than production.
• Temporal - a living current-state picture, maintained continuously by the mapping community, with each delivery reflecting the database as it stands at that sync. Retained prior cuts keep year-on-year comparisons measuring the world instead of a change in collection practice.
• Granularity - one row per property, as a geocoded node or a drawn footprint. No aggregates to unwind, no sampling weights to reverse-engineer; the 452,341st object is as legible as the first.

Against the wider catalog this record scores 8 out of 10 against a mean of 7.81 across 1,744 tracked datasets spanning 159 viable industries. Within the ten-record hotels, resorts & cruise lines pack it owns the property-level supply pole: official statistics count nights and receipts, this layer counts doors. See the whole ranking at best hotels-resorts-cruise-lines datasets.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and cadence; scoping, normalization and schema stability are our problem. The sixteen-column envelope above arrives unchanged through every channel - ids already stable for joining, addresses already flattened into clean components, brands resolved where tagged - so a competitive-distance matrix or a store-locator build starts on data rather than on parsing. Footprint elements come as centroid points or full polygons as your stack prefers, cadence changes are a settings conversation rather than a re-integration project, and large one-off backfills sit beside recurring syncs without special handling. A sample scoped to your cities and property families comes first either way.

Six workloads pay for themselves quickly.

• Site selection and whitespace mapping - lodging doors counted per drive-time, catchment or corridor on one taxonomy instead of stitched city directories.
• Chain-footprint tracking - operator and brand tags turn scattered pins into attributable networks, so rival coverage becomes a table rather than a hunch.
• Supply-density features - hotel counts and bed-place sums engineered straight into demand models and rate benchmarks as supply-side covariates.
• Enrichment and dedupe - commercial feeds cross-checked against a transparently documented global layer before they reach production dashboards.
• Opening and closure monitoring - consecutive retained cuts diffed into lead lists: what opened on the coast, what closed downtown.
• Finders and territory products - locator and coverage features standing on rows that carry boundaries as well as points.

Workflow detail lives at data scientists use cases and market researchers use cases.

Developers and data-product builders sit highest: one stable envelope with ids built for joining means locator, finder and territory features ship without an extraction project bolted on. Data scientists and ML engineers follow - supply-density features at planetary scale on a classification that behaves identically across borders. Competitive-intelligence teams get brand attribution that makes rival lodging footprints countable (developers builders use cases).

Market researchers and consultants gain reproducible supply numbers per territory, investors and quants gain the supply-side frame that aggregate demand series cannot provide, and journalists, academics and students gain a global inventory whose definitions anyone can read. The pack-wide pattern holds: registries and statistical offices go deep on their own estates, while this layer goes wide enough to frame them all.

Property-level supply reads differently depending on the neighbours. Official statistics frame demand and money: World Bank International Tourism Arrivals and Receipts count visitors and earnings by economy, Eurostat Nights Spent at Tourist Accommodation adds regional European depth, and Eurostat Capacity of Tourist Accommodation holds the official bed-stock counterweight. At booking grain, Hotel Booking Demand (Kaggle - 119k Portuguese Hotel Bookings) supplies the guest behaviour this inventory lacks. The head-to-head with annual earnings runs on the OpenStreetMap Hotels vs World Bank International Tourism Receipts comparison page, and the provenance story continues on the OpenStreetMap source profile. All ten, ranked and cross-linked, sit in the hotels, resorts & cruise lines data hub.