Extract POI data from OpenStreetMap: layers, fields, delivery

One mapped feature, one row. Every record carries four fixed columns - element type, a stable identifier, latitude, longitude - then the tag bundle volunteers attached: name, operator, brand, category, address components, phone, website, opening hours, wheelchair flag, whatever the mapper recorded. The envelope is deliberately boring so the useful part travels intact: identifiers already stable for joining consecutive deliveries, addresses flattened into clean components, geometry resolved to points or building footprints as your stack prefers.

Fill rates are the honest part. A name is close to universal and the contact fields run reliable; the capacity trio of stars, rooms and beds survives on only a small minority of hotel objects, so capacity analysis needs a declared fallback rather than a hopeful join. OpenStreetMap Hotels via Overpass Turbo documents all sixteen delivery columns with definitions verified against live records during the August 2026 cataloging pass, and OpenStreetMap Key:leisure — Global Leisure Features Wiki does the same for the leisure vocabulary - field dictionary first, rows second, brochure never.

Lodging is the deepest-documented consumer slice. At the August 2026 tag count the map holds 452,341 objects tagged tourism=hotel - 255,658 point nodes, 190,479 building-footprint ways and 6,204 relations - beside hundreds of thousands of hostels, guest houses and serviced apartments. One number deserves a warning label: just 329 objects carry an explicit resort tag, because most resorts simply wear the hotel tag. Filter on resort alone and you undercount the market by three orders of magnitude; the working cut is hotel-tagged properties screened by name and operator.

Sample rows lead every Datadory dataset page, because the fastest test of a POI layer is reading real records:

# one central-London hotel row, exactly as delivered - August 2026 research pass
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
internet   : wlan                    wheelchair : limited

Grain matters as much as count: one row per property, either a geocoded point or a drawn footprint, so catchments compute against buildings rather than dots wherever mappers drew them.

The leisure key is the widest lens: roughly 100 documented values across six families - green space, sports facilities, water venues, family recreation, relaxation and a long tail from bandstands to bird hides - attached to millions of mapped elements on every continent. Because the key attaches to nodes, ways and multipolygon areas alike, one schema spans a single picnic table and an entire country park.

OpenStreetMap Key:leisure — Global Leisure Features Wiki scores 10 on Datadory's rubric as the vocabulary record: every value defined, the companion keys (sport, brand, wheelchair, phone) documented beside it. Overpass API — OSM Leisure Query Engine, scored 9, is the serving half - fitness_centre, swimming_pool and pitch selections narrowed to any study area and answered with individual elements carrying all their tags, from tens of results to millions. Between them, a gym-density model or a pool-per-capita table starts as a request, not a research project.

Crowd-mapped density follows contributor populations, and we say so rather than average over it: strong across North America, Europe and Australia, thinner where fewer contributors have surveyed. Any count quoted per city needs that regional honesty attached.

The proof layer comes from official registers, pooled in the same fifteen-record specialized-consumer-services slice. In the UK, UK Business Register and Employment Survey (BRES) counts 2.73 million enterprises and 3.2 million local units across 28 tables per annual edition, broken down by 4-digit SIC class, region and employment size band. In the United States, U.S. Census Bureau Data API - CBP / Economic Census / Wholesale Trade serves County Business Patterns vintages 1986 through 2023 among 1,798 catalogued datasets, giving establishment, employment and payroll counts by county, place and ZIP code.

Join them deliberately: pull register counts for your geography, set the mapped POIs beside them, and read a large gap as a mapping-density signal - never as a market-size fact. The register says how many businesses trade; the map says which ones someone drew.

Two shapes fit two jobs, and Datadory delivers both under one contract.

Filtered cuts answer narrow questions fast: one tag family, one study area, tens of rows to millions. OpenStreetMap Overpass API, scored 9, runs the same selection model beyond leisure and lodging - highways, railways, pipelines, storage facilities - so a multi-layer territory file lands from one specification rather than five separate projects.

Regional snapshots suit repeat whole-country work. Geofabrik OpenStreetMap Data Extracts republishes the entire map as pre-clipped regional files: eight continents resolving to 555 named regions, Europe the heavyweight at 32.4 GB, Antarctica a 31.6 MB afterthought, roughly 78 GB across the top level. Clip polygons sit beside every region for cutting custom subsets, retained change history covers about 100 days of backfill, and weekly full-history archives allow a 2014-vintage network to be reconstructed exactly - the difference between a map and a longitudinal dataset.

Four personas account for most of the traffic, and they enter through different doors:

• Developers and data-product builders want locator, finder and whitespace features shipping without an extraction project bolted on - one envelope, stable identifiers, geometry pre-resolved.
• Data scientists engineer supply-density features straight into models - gym counts, pool provision, hotel supply per catchment - on a classification that behaves identically across borders.
• Competitive-intelligence teams turn brand tags into attributable networks, so rival footprints become a table rather than a hunch.
• Market researchers and investors need reproducible supply denominators: mapped POI counts per territory set against register counts, citable end to end.

The platform alternatives in the pool are excellent at what they expose and explicit about what they restrict. Fresha - Global Salon, Barbershop & Spa Marketplace pins 130,000-plus partner venues with starting prices and star ratings, but its terms confine reuse to its partner programme. Yelp's business-review corpus ships 6,990,280 reviews of 150,346 businesses across 11 US metropolitan areas, positioned for educational work.

The mapped layer is the counterweight: built for redistribution from the start, reaching every continent including the cities platforms skip, and carrying no ceiling on how much of it a product may serve. Through Datadory every record ships with its reuse terms stated up front - obligations arrive documented beside the rows instead of surfacing after launch - and the layers land joined on shared keys under one column contract.

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

A sample cut to your own categories and cities comes first either way - fifty real rows settle more arguments than any datasheet.

This page sits inside a broader landscape. The specialized consumer services data guide is the pillar: all fifteen pooled records scored and grouped by workflow. The specialized consumer services data hub indexes every dataset page with field dictionaries, coverage statements and quality scores, while the best specialized consumer services datasets ranking and the free specialized consumer services datasets shortlist sort the pool by different questions.

Sibling angles: what data exists on salons, barbershops and spas works the venue-and-price layer, and County Business Patterns through the Census Data API pairs mapped POIs with official establishment counts. The point of interest record glossary entry defines the row structure underneath everything above. For provenance beyond this slice, the Geofabrik source profile and the U.S. Census Bureau source profile cover what else these publishers ship.