Methodology

How a town gets its grade — and where the data runs out

TerraNova grades 51,317 municipalities across 4 countries from A++ to G, entirely from published official data. This page names every input, every source, and every place where a number is coarser than it looks.

Scoring model 2026.12-momentum-v1, last computed 2026-08-02. The tables below are generated from the live model and the live database, not written by hand, so they cannot drift out of step with the grades.

The one rule How a grade is built Whose priorities Every indicator How complete each grade is What we don't measure

The one rule: nothing rather than a plausible guess

The easiest way to build this site would be to fill every gap with something reasonable. A province's average price where a town's is unpublished. A national climate figure where the local one is missing. A zero where nothing was found. Every one of those produces a page that looks complete and quietly misleads, and a wrong number is worse than a blank one because you cannot tell it is wrong.

So, three commitments, which the rest of this page exists to let you verify:

  • A missing indicator scores nothing. It does not become a zero and it does not become an average. The pillar is recomputed over whatever inputs genuinely exist, and if none do, the pillar reads as “no data” instead of a middling score.
  • A proxy is labelled a proxy. Where the only published figure describes a province or a district rather than the town, we use it — a coarse number honestly labelled is still useful — and we say so, on the town's page and in the table below.
  • Absence is stated, not hidden. Where we have no data, the section is simply not there. A town page with fewer panels than another is telling you something true.

This is not an abstract principle. The bugs that hurt most in this project's history were all the same shape: a dead data source rendering a confident-looking score. An inland town once showed a tsunami risk bar. Half of France once carried an “attractiveness” score computed from map data that had never been imported for France — it read as a perfectly average result, which is exactly why it took so long to notice. Both are gone, and the rule above is what stops the next one.

How a grade is built

Four steps. Each indicator is normalised to 0–100. Indicators roll up into five pillars, each indicator weighted inside its pillar. The pillars combine into one 0–100 composite. The composite maps to a letter.

Normalisation is a percentile rank within the same country for most things: a price or a rental yield is scored against the rest of that country, because “expensive” only means anything relative to somewhere.

Two kinds of input are scored on an absolute scale instead, because ranking them would destroy the very thing they measure. Natural hazards carry a 30-year probability: ranking a hazard against a country where most towns have none compresses a real tenfold difference in exposure into nearly the same score. The population trend carries a percent per year: a town losing 1.2% of its people a year is doing the same thing in Galicia and in the Auvergne, and a rank would have said only “about average for its country” about both.

Pillar weights

The default (“balanced”) weighting. This is what every grade on the site means unless you changed the sliders yourself.

PillarWhat it answersWeight
Climate & risk Is the weather actually pleasant, and what could go wrong? 30%
Vitality Is this a living town with a future, or one that is emptying out? 25%
Services & access Can you live here without a car for everything, and get home from an airport? 20%
Value How expensive is it, per square metre, compared with the rest of the country? 10%
Market What have prices, sales and rents been doing? 15%

The letter

Fixed thresholds on the 0–100 composite, so a grade means the same thing in every country and in every report.

GradeComposite from
A++92
A+80
A68
B56
C44
D32
E22
F12
G0

A pillar with no data at all is treated as neutral (50) when the composite is formed, so that weights stay comparable between towns — and the town's page reports how many pillars that applied to, so a grade built on three pillars is never presented as if it were built on five.

Whose priorities? Three weightings, and yours

A retiree and a landlord are not looking for the same town, so one fixed ranking would be wrong for both. The pillar weights are adjustable, and the site ships three presets. Changing them re-weights the pillars — it never changes the underlying measurements, and it cannot flip an indicator's direction.

Profile Climate & riskVitalityServices & accessValueMarket
Balanced
The default, and what every grade on the site means unless you changed it.
30% 25% 20% 10% 15%
Lifestyle
For living there: climate and services matter more, market returns less.
35% 15% 25% 15% 10%
Investment
For letting or reselling: market and value dominate, climate comfort matters least.
15% 20% 15% 20% 30%

Every indicator, and where it comes from

Grouped by pillar. “Weight in pillar” is the indicator's share of that pillar, renormalised over whichever inputs a given town actually has. “Coverage” is the share of that country's municipalities for which the figure exists at all — the honest picture, measured from the database when this page was built.

Climate & risk

30% of the grade

Is the weather actually pleasant, and what could go wrong?

Everyday comfort

35% of this pillar · higher is better · ranked within its own country

How pleasant the year actually is — sunshine and mild temperatures, penalising both cold winters and punishing summer heat — from decades of reanalysis data at the town's own location.

Country Source Vintage Figure describes Coverage
ES Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
FR Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
IT Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
PT Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%

This is half the default climate pillar, and it is the biggest hole in the dataset: it has only been computed for a fraction of France and Italy so far. The coverage table below says exactly how much.

Beach season

35% of this pillar · higher is better · ranked within its own country · warm/coastal profile only

A long usable warm season plus a mild winter — the Mediterranean second-home question rather than the year-round-living one.

Country Source Vintage Figure describes Coverage
ES Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
FR Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
IT Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%
PT Copernicus Climate Change Service (C3S) — ERA5 2026 the town itself 100%

Deliberately not the inverse of everyday comfort: a place can score well on both. It is used only by the warm/coastal profile and never enters the default grade. Coverage is currently almost nil, so that profile falls back to the standard climate pillar.

Wildfire exposure

25% of this pillar · lower is better · absolute 30-year probability

The chance of a damaging wildfire near the town within the next 30 years, built from satellite fire detections around it.

Country Source Vintage Figure describes Coverage
ES NASA FIRMS — VIIRS active fire (fires near) 2026 the town itself 100%
FR NASA FIRMS — VIIRS active fire (fires near) 2026 the town itself 100%
IT NASA FIRMS — VIIRS active fire (fires near) 2026 the town itself 100%
PT NASA FIRMS — VIIRS active fire (fires near) 2026 the town itself 94%

An absolute probability, not a ranking. That matters: ranking hazards against a country where most towns have none would flatten a real tenfold difference into nearly the same score.

Earthquake exposure

10% of this pillar · lower is better · absolute 30-year probability

The chance of a damaging earthquake within 30 years, from the recorded earthquake catalogue around the town.

Country Source Vintage Figure describes Coverage
ES USGS FDSN — earthquake catalogue 2026 the town itself 100%
FR USGS FDSN — earthquake catalogue 2026 the town itself 100%
IT USGS FDSN — earthquake catalogue 2026 the town itself 100%
PT USGS FDSN — earthquake catalogue 2026 the town itself 94%

Also an absolute 30-year probability. Within the climate pillar, the worst hazard governs rather than the average — averaging lets an absent hazard mask a near-certain one.

Tsunami exposure

0% of this pillar · lower is better · absolute 30-year probability · not in the default grade

Recorded tsunami exposure near the town, on the same absolute 30-year scale as the other hazards.

Country Source Vintage Figure describes Coverage
ES NOAA NCEI/WDS — Global Historical Tsunami Database 2026 the town itself 100%
FR NOAA NCEI/WDS — Global Historical Tsunami Database 2026 the town itself 100%
IT NOAA NCEI/WDS — Global Historical Tsunami Database 2026 the town itself 100%
PT NOAA NCEI/WDS — Global Historical Tsunami Database 2026 the town itself 94%

Computed and shown on the town's page, but weighted zero in the default grade — re-weighting it would move every coastal ranking, and that decision has not been taken. It is a real number when you look at it, not a placeholder.

Vitality

25% of the grade

Is this a living town with a future, or one that is emptying out?

Population trend

40% of this pillar · higher is better · absolute % per year

How fast the town is gaining or losing registered residents, as a compound annual rate — the strongest single signal here, and the one that catches rural depopulation.

Country Source Vintage Figure describes Coverage
ES Annualised from the official population register 2025 the town itself 93%
FR Annualised from the official population register 2020 the town itself 97%
IT Annualised from the official population register 2025 the town itself 99%
PT Annualised from the official population register 2023 the town itself 99%

Unlike almost everything else on this page, this is an absolute rate rather than a ranking against its own country: −1.2%/yr means the same thing in Galicia and in the Auvergne. It measures the register, not the people, and a municipal boundary change reads as a collapse or a surge — which is why the ranked lists refuse to publish anything beyond ±5%/yr. Each town's own page states the years its figure was measured over, because the four statistics offices publish different windows.

Density

15% of this pillar · higher is better · ranked within its own country

People per square kilometre.

Country Source Vintage Figure describes Coverage
ES Derived from population ÷ surface area 2025 the town itself 99%
FR Derived from population ÷ surface area 2020 the town itself 100%
IT Derived from population ÷ surface area 2025 the town itself 100%
PT Derived from population ÷ surface area 2023 the town itself 99%

Derived from population and surface area rather than published as such. Municipal boundaries vary enormously in size between these four countries, so this reads as a rough town-versus-countryside signal, not a precise one.

Things to do

25% of this pillar · higher is better · ranked within its own country

Density of places worth going to — food, culture, leisure, tourism, nature — per square kilometre, counted from OpenStreetMap.

Country Source Vintage Figure describes Coverage
ES OpenStreetMap — points of interest 2026 the town itself 100%
FR OpenStreetMap — points of interest 2026 the town itself 100%
IT OpenStreetMap — points of interest 2026 the town itself 100%
PT OpenStreetMap — points of interest 2026 the town itself 100%

Only scored where the country's map data has actually been imported. Where it has not, this contributes nothing to the grade rather than counting as zero: a town is not empty just because we have not looked.

Foreign residents

20% of this pillar · higher is better · ranked within its own country

The share of residents who are foreign nationals — a proxy for how easy the landing is for an incoming foreign buyer.

Not measured identically everywhere: Spain, France and Italy count residents by nationality, while Portugal counts holders of a residence permit, which is a narrower group. Compare within a country more confidently than across them.

Services & access

20% of the grade

Can you live here without a car for everything, and get home from an airport?

Air connectivity

50% of this pillar · higher is better · ranked within its own country

How reachable international airports are, weighted by distance, so a town near one big hub and a town near three small ones score differently.

Country Source Vintage Figure describes Coverage
ES TerraNova — air connectivity (curated airports) 2026 the town itself 100%
FR TerraNova — air connectivity (curated airports) 2026 the town itself 100%
IT TerraNova — air connectivity (curated airports) 2026 the town itself 100%
PT TerraNova — air connectivity (curated airports) 2026 the town itself 100%

Everyday essentials

50% of this pillar · higher is better · ranked within its own country

How many of pharmacy, supermarket, school and doctor exist inside the town's own boundary, out of four.

Country Source Vintage Figure describes Coverage
ES OpenStreetMap — everyday essential services 2026 the town itself 100%
FR OpenStreetMap — everyday essential services 2026 the town itself 100%
IT OpenStreetMap — everyday essential services 2026 the town itself 100%
PT OpenStreetMap — everyday essential services 2026 the town itself 100%

Containment, not distance to the nearest one. Distance from a municipal centre is meaningless when the centre is an arbitrary point in a large rural boundary; "is there a pharmacy in this town at all" is a question with an answer.

Value

10% of the grade

How expensive is it, per square metre, compared with the rest of the country?

Price per m²

100% of this pillar · lower is better · ranked within its own country

The published price per square metre. Cheaper scores better — this pillar is affordability, not desirability.

Country Source Vintage Figure describes Coverage
ES MITMA — valor tasado de la vivienda libre (€/m²) 2026 3% the town itself, rest a wider-area average 99%
FR DVF — Demandes de valeurs foncières (Etalab) 2020 the town itself 95%
IT Agenzia Entrate — OMI, Quotazioni Immobiliari 2018 wider-area average (proxy) 92%
PT INE Portugal — valor mediano de venda €/m² (proxy distrital, interim) 2025 wider-area average (proxy) 100%

The single coarsest number on the site. For most of Spain and all of Portugal the only published figure is a provincial or district one, which is labelled as a proxy wherever it appears and is why no report compares your asking price against a local median in those countries.

Market

15% of the grade

What have prices, sales and rents been doing?

Price appreciation

30% of this pillar · higher is better · ranked within its own country

How fast prices have risen, annualised, over roughly the last five years.

Country Source Vintage Figure describes Coverage
ES INE — Índice de Precios de Vivienda (IPV, CAGR 5 ans par communauté autonome) 2025–2026 3% the town itself, rest a wider-area average 100%
FR DVF — Demandes de valeurs foncières (Etalab) 2023 the town itself 92%
PT INE Portugal — mercado imobiliário (proxy distrital, interim) 2025 wider-area average (proxy) 100%

Past movement, not a forecast. The window is not the same everywhere, so the town's own page states the years the figure was measured over rather than leaving "5y" in a label to imply it.

Market liquidity

25% of this pillar · higher is better · ranked within its own country

Sales per 1,000 dwellings — how readily property here actually changes hands, which is what determines whether you can sell again.

Gross rental yield

25% of this pillar · higher is better · ranked within its own country

A year of rent as a percentage of the purchase price.

Country Source Vintage Figure describes Coverage
ES SERPAVI (MIVAU) — precio de referencia del alquiler ÷ precio de compra (rendimiento bruto estimado) 2024 41% the town itself, rest a wider-area average 99%
FR ANIL "Carte des loyers" (loyers d’annonce) ÷ DVF (prix) — rendement locatif brut 2025 the town itself 94%
IT Agenzia Entrate — OMI, Quotazioni Immobiliari 2018 wider-area average (proxy) 87%
PT INE Portugal — mercado imobiliário (proxy distrital, interim) 2025 wider-area average (proxy) 100%

Gross, not net: no tax, no management, no vacancy, no maintenance. And it is one published figure divided by another, so it inherits the coarseness of both — where either side is a zone or province average, the page says so.

Tourism intensity

20% of this pillar · higher is better · ranked within its own country

Tourist bed places per 1,000 inhabitants — short-let demand, and equally a warning about summer crowding.

Counted differently in each country: registered tourist dwellings in Spain, measured bed places in Italy, and an equivalence that includes second homes in France, without which French resorts read as empty. Read it within a country.

Each source is linked to the body that publishes it, and the data remains theirs under their own licences — see the terms. Every town's own page repeats the source and date for each figure shown on it, so you can check a specific number rather than trusting this table.

How complete is each grade, really?

The uncomfortable table. For each country, the share of the pillar's intended indicator weight that has a real measurement behind it. 100% means every input the model wants exists; 50% means the pillar was computed from half of what it is supposed to weigh, and the letter is correspondingly less informative.

Country Climate & riskVitalityServices & accessValueMarket Whole grade In the dataset
Spain 100% 97% 100% 99% 99% 99% 8,217
France 100% 99% 100% 95% 94% 98% 34,888
Italy 100% 100% 100% 92% 64% 94% 7,904
Portugal 97% 99% 100% 100% 100% 99% 308

Read this before comparing a town in one country with a town in another. Right now Spain is the best-evidenced country in the dataset (99%) and Italy the least (94%), so the same letter is a stronger statement in one than the other. Where you see a figure below 50%, the pillar is running on a minority of its intended evidence and should be read as a hint rather than a finding.

This is deliberately a harsher measure than the completeness percentage shown next to a grade. That one counts how many of the five pillars have any data; this one counts how much of the evidence each pillar wanted actually arrived. A pillar surviving on a third of its intended weight counts as fully present in the first measure and as a third in this one.

What we measure but never score

Some figures are shown on a town's page and deliberately kept out of the grade, because scoring them would produce a ranking that is not valid. Showing them is useful; ranking on them is not.

Shown, not scoredWhy
PopulationA raw headcount is context, not quality — a big town is not a better town. It feeds density and the population trend, both of which are scored.
Published population changeThe raw percentage each statistics office publishes, shown on a town's page with the years it covers. Not scored directly: the four offices publish windows of 10, 5, 6 and 10 years under this one label, so the raw percentages are not the same quantity. It is annualised first, and the annual rate is what enters the grade.
Recorded burglary (France)Residential burglary per 1,000 dwellings. Not scored: crime statistics are published for large municipalities and withheld for small ones, so scoring them would systematically penalise exactly the small towns this product is for — and a withheld figure is not a zero.
Recorded crime (Portugal)All recorded crime per 1,000 inhabitants. Kept separate from the French figure on purpose: different offences, different denominator, different country. Collapsing them into one "safety" score would invite a comparison that is not valid.

What is not in a grade at all

The part you cannot infer from what is on the page — so it is written down.

Not measuredWhy not
Flood exposureWe have no usable municipal flood dataset covering these four countries. Flood risk is therefore absent from the climate pillar entirely rather than approximated from something adjacent. It would refine that pillar, not unlock it — the pillar already scores on comfort and two other hazards.
Air and water qualityPublished at station level, not municipal level, across all four countries. Interpolating a station reading to a town would produce a number that looks municipal and is not.
Schools, hospitals and transport qualityWe count whether a school or a doctor is present, not how good it is. Comparable quality measures do not exist across four countries, and inventing a composite from what does exist would be exactly the plausible guess this site refuses to make.
Anything about a specific propertyEvery grade describes a municipality. No building, no street, no view, no state of repair, no legal status. Two houses 200 metres apart get the same grade and can be worth very different money.

Grades move. That is the model improving, not the town changing

Official sources are revised and republished, new indicators land, and the scoring model changes when the data gets better. All three can move a town's letter without anything about the town having changed. Every score carries the model version that produced it (currently 2026.12-momentum-v1) and the date it was computed, and every paid report states both, so a document you downloaded last month can be read as the snapshot it is.

None of this is advice. It is a way to narrow tens of thousands of municipalities down to a shortlist worth visiting — see the terms for what that does and does not mean.

If a number looks wrong, tell us. Every town page carries a “something wrong on this page?” form. Corrections from residents and town halls are the one input we cannot get from an official dataset, and they are welcome.

Browse the data: the interactive map, the guides, or any town's page.