How Wandroz rates neighbourhood safety

2046 neighbourhoods across 65 cities in 22 countries. This page says where each rating's evidence comes from, what it can and cannot tell you, and where it is weak — so you can decide how much weight to give it.

How Wandroz assesses neighbourhoods

Where a city or country publishes neighbourhood-level crime data, Wandroz uses it. Where none exists — which is most of Europe — each area is instead reviewed individually against named local sources. Where neither is available yet, the city is labelled as such rather than dressed up as something stronger.

Every page on this site tells you which of those three it is on. They are never blended, and a rating is only ever meaningful relative to other areas of the same city: a green in Turin and a green in Berlin come from unrelated scales and are not comparable.

Administrative boundaries are official geometry in every city regardless of evidence class — real wards, districts, wijken or boroughs, not circles drawn around a point. It is the rating whose evidence differs, never the shape.

Evidence levels

How official-data assessments work

The shape is the same in every official-data city. Crime records published by the responsible authority are matched to that city's real administrative areas, using the authority's own boundaries rather than an approximation. Each area's recorded-crime level is then evaluated in the context of that city — never against a fixed international scale, and never against another city — and the result is expressed as a traveller-facing level rather than a number, because the underlying counts are not precise enough to justify one.

Where population is used to normalise a count, Wandroz applies additional safeguards for the cases where a resident-based rate is known to mislead — city centres with few residents and very high visitor footfall, above all. Where an area's rating carries that risk, the page says so in words rather than leaving the number to speak for itself.

Detailed scoring, weighting, classification thresholds and data-transformation logic are proprietary to Wandroz and are not published. Everything needed to judge the evidence — the source, its year, the geography, and the limitations — is on this page and on each area's own page.

How local-source assessments work

This is deliberate, documented work per area, and it is not a police dataset, which is why it carries a different label from one. The principles it follows:

• Official data is checked for first. Where a city or country publishes an open, geolocated, neighbourhood-level crime dataset, the city moves to official-data class and this method is not used.
• Attributable sources. Established local and national news outlets, municipal and government publications, police-published district figures where they exist, official reports, and credible survey or research data. Where a source is a resident-perception survey or a market ranking rather than a crime figure, the area's page says so.
• Area by area, never city by city. A city-wide reputation is never copied across its districts. Each of the 1075 areas has its own review and its own written reasoning.
• Traveller-relevant signals. Theft and pickpocketing, robbery, violent incidents, drug-related activity, persistent disorder, nightlife trouble, transport-hub problems — things that would change how a visitor uses an area.
• Cross-source consolidation. One article or one isolated incident is not a rating. Where sources conflict, both findings are stated on the area's page rather than averaged away.
• Recency. Recent evidence counts for more; older material is used where it establishes a persistent pattern rather than a one-off.
• No inference from who lives there. Ratings are not derived from income, demographics, housing type or general reputation.
• Absence of reporting is not evidence of safety. See below — this is the rule the method lives or dies on.

Important limitations

City data sources

Counted from the published site at build time, including the unflattering figures. "Areas" is how many neighbourhoods that city has on Wandroz; "Period" is the reference period of the evidence behind the rating.

Source names link to the dataset or publication the rating is built on, where the publisher makes one reachable. For local-source cities the "Period" column gives the date the area-by-area review was completed; the individual sources behind each rating are named on that area's own page, which is where the evidence for a specific neighbourhood should be checked.

Local-source coverage, city by city

A high "no area-specific findings" count means the review reached fewer sources in that city — usually because its official geography is cut into small administrative units — not that the city is safe.

What a rating represents

A Wandroz rating is a traveller-facing reading of an area's recorded-crime level or documented safety signals, relative to the rest of that same city. It is not a score, not a probability, and not a prediction. Three levels are used because the underlying evidence does not support more precision than that, and inventing precision the data cannot carry would be the easiest way to make this site look authoritative and be wrong.

Detailed scoring, weighting, classification thresholds and data-transformation logic are proprietary to Wandroz and are not published. The evidence is public: the sources, their periods, the geography and the limitations are all on this page, and the reasoning for an individual area is on that area's page. What is not public is the recipe for turning one into the other.

Corrections

If a rating looks wrong, a source is misattributed, or a boundary is off, that is worth knowing and worth fixing. Write to hellowandroz@gmail.com naming the city and the area, and, where you have one, the source that contradicts what is published. Corrections that stand up are applied at the next build, and every build goes through the same automated checks as any other change.

London's scores are recomputed from data.police.uk by pipeline rather than by hand, and every borough page names the exact months behind its score — currently May–July 2026. That refresh runs on review rather than on an unattended schedule, so the window published here is read from the data itself and cannot promise a freshness the site does not have.

← Back to the city index