Methodology
GPM is a data engine before it is a website. This page describes what is collected, how it is turned into comparable metrics, and the rules that decide whether a market is allowed to publish at all.
Status of this dataset
Official national indicators and explicitly scoped city benchmarks now use BIS, OECD and national-statistics data covering 66 countries and 685,964 country-quarter observations. City prices, rents, yields and listing inventory remain clearly labelled synthetic prototypes while production listing connectors are built.
Pipeline
From source to page
The same six stages run for every country. Launching a new market is a configuration change, not a new codebase.
01
Source registry
Every country has a registry of connectors: listing portals, statistical agencies, central-bank series, cadastral extracts and developer feeds. Each one carries its own cadence, reliability weight and coverage note, and each is monitored for record counts, schema drift and field-null rates.
02
Ingestion
Scheduled jobs collect listings and public series. The raw payload is written before anything is transformed, so a normalization bug never destroys the underlying record and history can be rebuilt from the archive.
03
Normalization
Currency is converted at the observation date, area is standardised to square metres, geography is resolved against the hierarchy, property types are classified, obviously invalid values are dropped, and cross-source duplicates are collapsed onto a single listing identity.
04
Observation history
Prices are appended, never overwritten. A listing that falls from $300,000 to $285,000 keeps both observations, which is what makes reductions, days on market, listing velocity and genuine historical series possible later.
05
Metric computation
Metrics are recomputed per geography, asset class and period. Each metric carries its own minimum sample before it may be published, so a market can publish a price without publishing days on market.
06
Quality control
Before anything is served, the run is checked for sample collapse, implausible median movement, source volume drops, currency errors and duplicate spikes. A failed check withdraws the affected pages rather than publishing a number we cannot defend.
Definitions
What each metric counts
Every metric has a minimum sample of its own. Below that floor the metric is withheld even when the rest of the page publishes.
| Metric | Definition | Unit | Min sample |
|---|---|---|---|
| Median asking price per m²median_price_per_sqm | Median advertised sale price divided by advertised interior area, converted to USD at the month-end rate. | USD per m² | 30 |
| Median asking pricemedian_price | Median advertised sale price across qualifying listings, converted to USD. | USD | 30 |
| Median monthly asking rentmedian_rent | Median advertised long-let rent per calendar month. Short-let and seasonal advertisements are excluded. | USD per month | 30 |
| Gross rental yieldgross_yield | Annualised median rent divided by median asking price for the same geography and period. Gross of costs, taxes and vacancy. | Percent | 40 |
| Active inventoryactive_inventory | Distinct listings seen in the last refresh after deduplication across sources. | Count | 20 |
| New listings, 30 daysnew_listings_30d | Listings observed for the first time in the last 30 days. | Count | 20 |
| Removed listings, 30 daysremoved_listings_30d | Listings that stopped appearing in source refreshes. Removal is not the same as a completed sale. | Count | 20 |
| Listings with a price reductionprice_reduction_share | Share of active listings whose observed price fell at least once in the last 90 days. | Percent | 60 |
| Median days on marketmedian_days_on_market | Median days between a listing's first observation and its removal, for listings removed in the last 90 days. | Days | 50 |
Quality control
Publication rules
A geography exists in the database long before it deserves a URL. These are the checks a market has to pass, evaluated on every run.
Qualifying listings
at least 30
Below this the sample is too small for a stable median.
Reliable sources
at least 1
A source counts once its confidence weight reaches 75%.
Listing activity, 30 days
at least 12
Proof the market is still refreshing rather than sitting on stale rows.
Metrics clearing their floor
at least 3
A page with one number is not a market page.
Internal links
at least 2
A market has to sit inside the graph, not dangle off it.
Parent geography
published
A city cannot outrank the country it belongs to.
What happens when a check fails
The market is withheld: no page, no sitemap entry, no API response, and no place in any ranking. It keeps ingesting in the background and republishes automatically once it clears. Currently 2 of 25 tracked markets are withheld, and they are listed openly on the coverage page.
Anomaly review
Movements beyond 15% year over year on price, 20% on inventory, or a reduction share above 25% are flagged on the market page for human review. Flagging does not hide the number; it marks it as worth checking against the source before it is quoted.
Known limits
- Asking prices are not transaction prices, and they lead recorded sales.
- A removed listing is not proof of a sale.
- Gross yield ignores costs, taxes, vacancy and financing.
- Advertised area definitions differ between markets and sources.
Provenance
Source registry
Production and planned connectors, followed by the synthetic source shapes used to exercise listing-depth pages while live integrations are built.
Official global connectors
Selected residential property prices
active · quarterly
Bank for International Settlements
BIS-selected national series chosen for cross-country comparability. Individual national compilation methods still vary.
Analytical house price indicators
active · quarterly
Organisation for Economic Co-operation and Development
Complements BIS with rent and valuation ratios and provides overlapping house-price series for reconciliation.
Detailed residential property prices
active · quarterly
Bank for International Settlements
Extends country coverage with explicitly reviewed whole-country series from the heterogeneous BIS detailed dataflow.
Detailed residential property prices — reviewed city benchmarks
active · quarterly
Bank for International Settlements
Publishes current official city benchmarks when no qualifying national series exists, without presenting them as national coverage.
Prices of dwellings in new residential buildings
active · quarterly
Statistical Office of Montenegro
Discovers and archives the latest official MONSTAT workbook, publishing only the nationwide new-dwelling transaction series.
Residential Property Price Index
active · quarterly
National Statistics Office of Georgia
Uses Geostat’s current machine-readable workbook and preserves its Tbilisi-only, new-dwelling, offer-price scope.
Residential Property Price Index
active · quarterly
Statistics Mauritius
Discovers the current official historical-series workbook and publishes the comparable nationwide transaction-price index.
Residential Property Price Index
active · quarterly
Central Bank of Armenia
Uses the Central Bank’s live full-history XLSX export and publishes the nationwide all-residential-property index.
Housing market price index
active · quarterly
National Statistics Committee of Uzbekistan
Downloads the official SIAT CSV and chain-links the published nationwide quarterly relatives into a transparent normalized index.
Housing Price Index
active · monthly
National Statistics Office of Mongolia
Queries the official PxWeb API and publishes the current updated-model series with explicit Ulaanbaatar scope.
DVF residential sales and ANIL rent estimates by commune
active · annual
French public finances directorate (DGFiP) and ANIL
Streams and aggregates non-identifying 2025 DVF transactions, then joins ANIL rent estimates by INSEE commune code. Publishes derived statistics only; ANIL reuse requires attribution.
Local residential sale prices and average rents
active · quarterly
Central Statistics Office Ireland and Residential Tenancies Board
Pairs CSO HPM07 execution medians with CSO-hosted RTB RIQ02 average rents for 26 exactly matched geographies. Yield is explicitly labelled as a median-price/average-rent proxy.
Existing own homes — average purchase prices by region
active · annual
Statistics Netherlands (CBS) and Kadaster
Uses definitive notarial transaction averages from CBS StatLine. These composition-sensitive averages are explicitly not presented as constant-quality house-price indices.
Local house sales and rental statistics
active · quarterly
Statistics Portugal
Pairs current official sale and new-lease rent indicators at municipality level and derives a transparent gross yield from matched per-square-metre medians.
UK House Price Index — local authorities
active · monthly
HM Land Registry
Queries the official linked-data API for major city local authorities and London boroughs under the Open Government Licence.
Residential Property Price Index
active · quarterly
Kosovo Agency of Statistics
Parses the full official index table from the latest quality-qualified KAS release and preserves its nationwide transaction scope.
Listing-depth prototype sources
Colombian listing portals (aggregated)
CO · daily
Residential sale and rental advertisements across the six largest metros.
Confidence 86%
Colombia national statistics office
CO · quarterly
Building permits, construction cost and dwelling stock series.
Confidence 95%
Colombia central bank housing series
CO · quarterly
Used-dwelling price index used to sanity-check asking-price drift.
Confidence 97%
Mexican listing portals (aggregated)
MX · daily
Sale and rental advertisements weighted toward metro and resort markets.
Confidence 84%
Mexico cadastral and registry extracts
MX · monthly
Parcel geometry and land use, used for geocoding and land classification.
Confidence 90%
Mexico developer inventory feeds
MX · weekly
New-build release schedules in Quintana Roo and Yucatán.
Confidence 79%
Spanish listing portals (aggregated)
ES · daily
Sale and long-let advertisements across provincial capitals and coastal markets.
Confidence 89%
Spain national statistics office
ES · quarterly
Transaction counts and registered sale prices by province.
Confidence 96%
Portuguese listing portals (aggregated)
PT · daily
Sale and rental advertisements for mainland municipalities and the Algarve.
Confidence 87%
Portugal national statistics office
PT · quarterly
Municipal median sale values per m² and dwelling transaction counts.
Confidence 96%
Access
Read the same numbers as JSON
The API serves the published set only, from the same read model the pages use.
- /api/global-baseline
66 official national property-price series.
- /api/markets
Published countries and cities with headline metrics.
- /api/markets/colombia/medellin
A single market, with history and sources.
- /api/rankings
Every ranking table, or one by slug.
Official source data through 2026-Q3 · ingested Aug 9, 2026 · city prototype through Jul 31, 2026