Sources, scores and limitations

What this dataset represents

Version 2026-09-dubai · last editorial update 2026-09-09. Coverage: Georgia, Spain, Portugal, Greece, France, Germany, United Arab Emirates. We start with a limited set that can be researched responsibly. Coverage is not a claim to include every attractive property market.

Each observation stores its value, unit, original currency where relevant, geographic scope, reference period, source publication date when known, review date, confidence and methodology. Country profiles separate sourced rules from SafeBuy’s editorial interpretation. Unknown data remains unknown.

Government, statistical, tax and immigration sources are preferred for rules and official series. Original market research can be used for asking-price and rental samples. These are different measurements: national transaction indices measure change, listing benchmarks measure advertised prices, and neither proves the value of a specific property.

Transparent scoring

Comparison rules 2.0; editorial anchor configuration 2.0. For each selected dimension, a published value is scaled to 0–100 using fixed anchors. Values outside the anchors are clipped. For a higher-is-better dimension: 100 × (value − minimum) ÷ (maximum − minimum). A lower-is-better dimension reverses that result.

The default explorer shows published benchmarks without a composite score. Explore markets sorts alphabetically; the price and rental views sort the indicated observation, keeping missing observations visible separately. Source geography and methods still differ, so this order is not an investment recommendation.

For a selected comparison of two or three markets, an advanced score requires every active dimension to have current, sourced evidence in every selected market. Each dimension must also share the same nonempty editorial comparisonGroup across those markets. The final score is the sum of each normalized value multiplied by its selected weight, divided by the sum of all selected weights. Missing, low-confidence, stale or differently measured evidence prevents the complete comparison score. No market or priority is silently dropped. This rule applies even when an older published configuration retains a lower minimumCoverage setting.

Buying-cost priorities are disabled while no usable total-cost observations are available. Registered transaction prices and asking prices do not share a comparison group simply because they can be expressed in the same currency. No market receives a commercial score bonus, and historical growth is not a forecast.

EU-only filters country membership. Financing, residence interests and rental strategy add relevant checks, without inventing a legal-risk, liquidity or bank-eligibility score. Nationality is not collected to imply a legal eligibility decision. Tax sensitivity, land restrictions and currency exposure must be assessed for the actual investor and property.

Yields, costs and budgets

Gross rental yield is annualised rent divided by a purchase-price benchmark, multiplied by 100. Read the original methodology: the units and locations sampled vary. It excludes vacancy, repairs, management, service charges, insurance, tax and debt payments. We do not invent net yields.

Georgia’s price and gross-yield observations are read from the existing SafeBuy Georgia Property Price Index, with its permanent monthly citation. The adapter supplies separate Tbilisi and Batumi observations, identifying each city and sample size. The legacy Georgia country metric continues to reference Tbilisi for compatibility; it is not a nationwide average. It never copies a price into an independent international table that can drift from the index.

Purchase-tax rules are not automatically total acquisition costs. Total-cost scores require a sourced all-in estimate; otherwise the facts remain in the country profile. The budget illustration divides your budget by the cited price benchmark. It excludes taxes and purchase costs, does not reserve cash, and does not assert that a property of that size is available. Budget changes the floor-area illustration; it does not mechanically change country scores.

Purchase scenarios and shared comparisons

The optional scenario is for a cash purchase of a completed apartment with long-term tenants. Each selected market has independent visitor-entered values. Unknown expenses remain blank; an explicit zero means the visitor chose to assume no cost. Partial subtotals are labelled and complete returns are withheld until the required inputs are complete.

Collected rent equals monthly rent multiplied by (12 minus vacant months). Management fees apply to collected rent. Operating surplus subtracts management, annual service charges and maintenance. Initial cash includes the purchase, itemized acquisition costs, furnishing and a separate reserve. The cash return divides operating surplus by that initial cash. A reserve is not charged again as an annual expense. Outputs are before personal tax, financing and resale costs.

The editable stress example initially reduces rent by 10% and adds one vacant month, capped at twelve. These are scenario assumptions, not market evidence. Market benchmarks never establish the rent of a specific apartment or expected return.

Copy comparison link creates an ungated URL containing selected markets and optional numeric assumptions. It contains no contact details. Anyone with the link can read those assumptions. Shared comparisons load the latest evidence; removed markets are disclosed. Print/save as PDF records the currently displayed evidence and assumptions. Personalization does not create additional indexable pages.

Currency and freshness

Original currencies are preserved. EUR is the comparison basis for price scoring. GBP, USD and AED comparisons use the dated FX observations: European Central Bank reference exchange rates, 2026-10-30. Operational FX is refreshed daily from ECB and CBUAE sources independently of editorial publication. CBUAE figures are published as reference rates for VAT obligations; their use here is labelled indicative research conversion. Rates are currency units per EUR, with per-currency dates when sources differ. Conversion is value ÷ source-currency rate × destination-currency rate. These are research reference rates, not live transaction quotes.

A failed source refresh retains its last validated dated observation. Cross-currency conversion stops when either required rate is older than 14 days; same-currency values remain usable. A missing currency rate stays unavailable. No default 1:1 conversion is used.

Every metric has an explicit age limit based on the older of its observation date and review date. At 80% of that limit it is marked “review due soon”; beyond the limit it is excluded from scores and data-led metric rankings. Historical figures remain visible in profiles, clearly marked stale. Rechecking an old observation cannot make it new. Rules are flagged using their review dates and require human review.

Editorial process and commercial relationship

Authorised SafeBuy editors maintain the source-linked dataset through the existing admin system. Publication validates references, fields, dates and scoring configuration. Updates retain a revision history. New country or comparison pages need useful evidence and a distinct research purpose; we do not publish every country pair or arbitrary budget variant.

SafeBuy’s commercial services focus on Georgia. Listings, valuation, registry checks and a call with David Lee are offered when you want to explore that market. No phone number is required to view the research. You can choose whether your callback request includes your shortlist and budget. Read how SafeBuy is paid.

To request a correction, use the contact form and include the country, observation and primary source. French editorial content can be reviewed separately using language-independent country IDs and metric fields; no unreviewed translated pages are automatically indexed.