Resources

10 datasets every real estate professional should know (+1).

Free building and city datasets, with links, licences and the catch for each one. Checked on 4 November 2026.

Most building datasets tell you what exists. Almost none tell you what is being built.

We went through the free datasets real estate teams actually reach for and wrote down the link, the licence and the one limitation to keep in mind for each. Licence is the sleeper issue for professional use, so it gets its own line every time.

Quick licence scan before you start

Non-commercial: the main Global Building Atlas layers (CC BY-NC 4.0). Check before using them in a paid product.
Share-alike (ODbL): OpenStreetMap and the Overture buildings layer. If you build a derived database and distribute it, it has to stay under the same licence.
Attribution required: Inside Airbnb (CC BY 4.0), GHSL (acknowledge the source).
Public domain: the Dutch BAG (Public Domain Mark 1.0).
Mixed or per file: EUBUCCO, Eurostat downloads, Kaggle (every dataset has its own label, and re-uploads are sometimes wrong).

Datasets, honest limits

OpenStreetMap

A volunteer-built, editable world map that includes buildings, roads, shops, transit stops and land use.

Linkopenstreetmap.org, bulk data at planet.openstreetmap.org and regional extracts at download.geofabrik.de CoverageGlobal, but building completeness is very uneven. A Nature Communications study of 13,189 urban agglomerations found more than 80% building completeness in 1,848 of them (16% of the urban population) and below 20% in 9,163 cities (48% of the urban population). Paper, summary by HeiGIT. UpdatesEdited continuously, the full planet file is republished weekly. LicenceODbL. Credit OpenStreetMap, and a derived database may be distributed solely under the same licence. Copyright page Use it forCounting shops, cafes, transit stops and schools within walking distance of a site to score a neighbourhood, using a free regional extract. The catchCompleteness depends on whether local volunteers mapped the area, so a thin result can mean an unmapped city rather than an empty one.

Overture Maps

An open global map dataset backed by large tech companies. The buildings layer merges OpenStreetMap, Esri Community Maps, national sources and machine-learning footprints from Microsoft and Google into one file.

Linkdocs.overturemaps.org, buildings guide CoverageGlobal. The 2026-10-21.0 release lists 2,533,842,612 buildings (about 2.5 billion), of which Google 898M, OpenStreetMap 707M and Microsoft 689M by source. Release notes UpdatesMonthly releases. LicenceThe buildings theme is ODbL because OpenStreetMap is the top-priority source. Per-source licences include CC BY 4.0 (Google, Esri). Attribution and licences Use it forQuerying one consistent building layer across several countries with a single tool (GeoParquet on S3 or Azure) to size and compare districts without stitching national files together. The catchEach footprint carries the age of its original source, so a new monthly release does not mean the buildings inside are new. The October 2026 release, for example, re-promoted September's buildings data with 0.00% change.

Global Building Atlas (TU Munich)

A global set of building outlines with an estimated height for almost every one.

LinkGitHub, data on mediaTUM, paper in ESSD, TUM press release CoverageGlobal, 2.75 billion building polygons, 2.68 billion with LoD1 (simple 3D block) models, and a 3 m height map. LoD1 height error (RMSE) is 1.5 to 8.9 m depending on the continent. UpdatesA single snapshot, built mainly from satellite imagery of 2019. LicenceNON-COMMERCIAL. GBA.Polygon, GBA.LoD1 and GBA.Height are CC BY-NC 4.0. A GBA.ODbLPolygon part is ODbL, and the code is MIT with Commons Clause. Use it forMultiplying footprint by height to estimate built volume and floor area per district, for density comparisons or a rough redevelopment screen. The catchThe headline layers carry a non-commercial licence and are a 2019 snapshot. Check the licence before using it in a paid product, and do not read it as current.

Microsoft Global ML Building Footprints

Building outlines detected by machine learning from satellite and aerial imagery, published as downloadable files.

Linkgithub.com/microsoft/GlobalMLBuildingFootprints CoverageWorldwide, 30,340 tiles across 225 regions. The repository states 1.4B buildings detected from Bing Maps imagery between 2014 and 2024, while its own FAQ gives 999M polygons, so read the numbers with care. Some areas are missing. UpdatesIrregular. The latest changelog entry (2026-08-13) refreshed the global dataset and updated 1,945 tiles. Later updates add Vexcel imagery from 2021 to 2025. LicenceCDLA Permissive 2.0 per the repository today. Some third-party pages still list ODbL, which likely reflects the older terms, so read the repository. Use it forFilling in building footprints in countries without an open cadastre, for example to estimate the number and size of buildings in a new market. The catchNeighbouring buildings can come from images up to a decade apart, so presence in the dataset says little about when a building appeared.

EUBUCCO

A harmonised database of European buildings that combines government registers, OpenStreetMap and Microsoft footprints, with height, floors, use and age where available.

Linkeubucco.com, documentation, v0.1 record on Zenodo Coverage322M+ buildings in 30 countries (EU-27, UK, Norway, Switzerland), from 55 sources: 62.2% government registries, 20.4% Microsoft, 17.4% OpenStreetMap. UpdatesVersioned releases. The initial public version (v0.1, October 2022) is described as a beta. LicenceMostly ODbL. The v0.1 record says 95%+ of the data is ODbL, with some inputs under other terms (for example Prague CC-BY-SA and Abruzzo CC-BY-NC), and warns that the database combines files under different licences. We could not find a licence statement on the current site, so check per file. Use it forComparing building stock between European cities by residential and non-residential share and size, before paying for local data. The catchAttributes are often estimated. Per the docs, just 43.2% of heights and 15.6% of construction years are original ground truth, and 56.7% of heights are ML-estimated. Age is the thinnest attribute.

GHSL (Global Human Settlement Layer)

European Commission grids showing built-up surface, building height, population and degree of urbanisation for the whole planet.

Linkhuman-settlement.emergency.copernicus.eu, datasets, GHS-BUILT-S R2023A CoverageGlobal. Built-up surface, population and degree-of-urbanisation grids list epochs from 1975 to 2030 (release R2023A) at 100 m or 1 km depending on the product. A separate 10 m built-up version exists for 2018. The building height product covers 2018 alone. UpdatesPeriodic releases, not continuous. LicenceThe datasets page says reuse is authorised with proper acknowledgment of the source. We could not confirm a formal licence name. Use it forComparing how fast built-up area grew around different urban regions since 1975, to see which markets are expanding outward. The catchCells are 100 m or coarser and epochs are five years apart, so it shows long-run growth rather than a single new project.

Dutch BAG (the example of a national building register)

The official Dutch register of every address and building, maintained by municipalities and distributed nationally by the Kadaster.

Linkpdok.nl, PDOK BAG page, download service, OGC service, Kadaster BAG CoverageThe Netherlands. PDOK offers buildings (panden), the residential and commercial units inside them, and mooring and standing places. That PDOK product includes main addresses alone. UpdatesThe PDOK service is updated daily. LicenceOpen data under the Creative Commons Public Domain Mark 1.0, per the Kadaster. Use it forLooking up a building and its units, usage purpose and floor area, and using the status field to screen buildings that are permitted or under construction. The catchStatus depends on municipal updates and can lag reality. Almost no other country publishes a register this complete.

Inside Airbnb

Independent, publicly scraped Airbnb listings, reviews and calendars for selected cities.

Linkinsideairbnb.com/get-the-data, data assumptions CoverageDozens of cities worldwide (at least 64 visible on the page), with country files for Australia, Canada, France, Germany, Greece, Italy, the Netherlands, Portugal, Spain, Sweden, the UK and the US. UpdatesQuarterly data for the last year is free. Each city shows its snapshot date. LicenceCreative Commons Attribution 4.0. Use it forMapping short-term rental density and listing prices by neighbourhood to judge how exposed a residential area is to Airbnb conversion. The catchIt is scraped, not Airbnb's own data, and occupancy and income are modelled on assumptions (a 50% review rate, a 3-night default stay, a 70% occupancy cap).

Eurostat / GISCO census population grid 2021

Official EU population counted per 1 km square from the 2021 censuses, with 13 census variables such as age and employment status.

LinkEurostat GISCO population grids CoverageEU-wide 1 km grid. Older GEOSTAT 1 km grids exist for 2018, 2011 and 2006, and derived products include a 1 km dwelling grid (Release 10/2025) and a 100 m population grid (Release 07/2026). UpdatesCensus-based. The 2021 grid has been reissued four times, latest version 3 on 30 May 2026. LicenceNot stated on the page. Downloads are subject to download rules you accept, with a link to Eurostat's copyright notice. Read both before commercial use. Use it forCounting how many residents live within 1 km of a retail or residential site, with the same method in every EU country. The catchA census snapshot, so it cannot show population change since 2021, and 1 km cells are coarse for dense city blocks.

Kaggle (and how to find more there)

A hosted library of user-uploaded and publisher-uploaded datasets, free to download with an account.

Linkkaggle.com/datasets CoverageCoverage, updates and licence vary per dataset. See the checked list below. Use it forPrototyping models and dashboards quickly on tidy sample data (house sales, Airbnb, permits) before paying for or collecting production data. The catchMany datasets are frozen copies from 2016 to 2020, some are scraped listings with unclear rights, and licence labels on re-uploads can be wrong.

Twelve real estate datasets we checked

"Updated" is the date of the latest upload on Kaggle, not the date of the data inside. Metadata read on 4 November 2026.

#	Dataset	What it covers	Updated	Licence	Watch out
1	NYC Property Sales	Every building or unit sold in NYC over one 12-month period: location, type, price, date	2017	CC0	Stale mirror of the NYC Department of Finance rolling sales. Current official files cover Oct 2025 to Sep 2026
2	House Sales in King County, USA	Seattle area home sales, May 2014 to May 2015	2016	CC0	A teaching dataset for regression. Not a current market view
3	Melbourne Housing Market	Melbourne housing clearance data from Jan 2016	2018	CC BY-NC-SA 4.0	Stale and non-commercial
4	Urban House Prices in Europe	Price per m2 for 50 European cities, plus GDP per capita and population density	2024	CC0	Most recent European city-level set we found, but no source is given, so check the city list and provenance
5	Apartment rental offers in Germany	Rental listings scraped from ImmoScout24 (2018 to 2019 in the original)	2020	(c) Original Authors	Scraped from a portal, listed as research use. A usability score of 1.0 does not mean current
6	Chicago Building Permits	Valid permits issued by the City of Chicago from 2006	2019	CC0	Stale mirror. Use the live city portal
7	San Francisco Building Permits	About 200k permits over 5 years (Jan 2013 to Feb 2018)	2018	ODbL	Frozen, good for practice
8	US Building Permits: 16 Cities, Sep 2026	43,119 permits across 16 US cities in the 30 days to 2026-09-22, one schema	2026	CC BY 4.0	Freshest permit set we found, but a third-party aggregator with a one-month window, large cities partly sampled, usability 0.59. Go to the city portals for history
9	Madrid real estate market	Listings crawled from Madrid property portals (21,742 rows, 58 columns)	2020	(c) Original Authors	Scraped portal listings, rights unclear
10	U.S. Airbnb Open Data	Several Inside Airbnb city files in one CSV	2023	CC0	Mirror of Inside Airbnb, which is CC BY 4.0. A CC0 label on a re-upload does not remove the attribution duty. Go to Inside Airbnb
11	New York City Airbnb Open Data	NYC Airbnb listings (2019)	2019	CC0	Hugely popular but stale. Same mirror caveat as 10
12	Zillow Economics Data	Zillow Home Value Index and Zillow Rent Index series by region	2018	(c) Original Authors	Stale mirror. The description points to zillow.com/data for current data

Also seen in search, checked at listing level alone: Seattle Building Permits (2018, CC0) and Los Angeles Building and Safety Permits (2019).

What the check showed: 9 of the 12 were last updated between 2016 and 2020. The four most-voted sets (NYC Airbnb, King County, Melbourne, Zillow) were all last updated between 2016 and 2019. The one set from this year is a low-usage aggregator. Usability scores measure documentation completeness, not freshness: the 2020 German rentals set scores 1.0, the freshest permit set scores 0.59.

How to search Kaggle without wasting a day

Search by pattern. Try "city building permits", "city property sales", "city airbnb", "country house prices", "country rent listings". Add the native word if the data is not in English (immobilien, vivienda, woningmarkt).
Sort on both axes. Sort by most votes to find well-tested sets, then re-sort by most recently updated. Check the "last updated" date on the dataset page, and remember it is the upload date. Open the data tab and look at the actual date column.
Check the licence on every dataset page. Treat CC0 on a re-upload with suspicion if the original is CC BY, and treat "(c) Original Authors" or "research use" as a stop sign for commercial work.
Do not trust the usability score as a freshness or quality signal. It reflects metadata completeness.
Prefer datasets that name their source, then go to that source. Airbnb: Inside Airbnb. Permits: the city open-data portal. NYC sales: the NYC Department of Finance. Prices and rents: the statistics office. Use Kaggle to learn the schema, then pull the current file from the original.
Read the notebooks (the Code tab) for the cleaning steps others had to do. For scraped listings, remember that portals often forbid reuse, so a popular notebook is not a licence.
Command line: kaggle datasets list -s "building permits" --sort-by updated runs the same search from a terminal (check kaggle datasets list -h for the exact flags).

Urban Construction Monitor

Every dataset above tells you what exists. This one is about what is being built right now. Monthly construction activity across whole cities, at parcel level, with history back to January 2018, built from satellite data. Live for Amsterdam, Utrecht, Den Haag and Eindhoven.

Linkurbanmonitor.eu UpdatesMonthly. LicenceCommercial service by Geospatial Data Consulting. Exports on licensed plans. Use it forSeeing where building work is actually active in a city this month, for pipeline tracking, competitor supply and site-neighbourhood checks. The catchLive today for four Dutch cities. Other cities are added on request, not yet in the app.

Request access Explore the maps

Also worth knowing

Google Open Buildings covers Africa, South and Southeast Asia, Latin America and the Caribbean, not Europe or North America. 1.8 billion detections, CC BY 4.0 or ODbL. sites.research.google/gr/open-buildings

Questions and corrections

Spot something out of date, or a dataset we should add? Write to hello@geospatialdataconsulting.com. Licences and release numbers change, so always check the source page before you build on anything here.

Written by Milan Janosov, The New Science of Maps. Dataset figures and licence labels are as read on the linked pages on 4 November 2026.