How the scores work
Every report scores an address from 0 to 10, with a plain label next to each number. This page explains where each number comes from, and what it can and can't tell you.
The scale
Each score is shown with a number and a label. The label comes from the number:
| Score | Label |
|---|---|
| 8 to 10 | 8.0 Excellent |
| 6.5 to under 8 | 6.5 Good |
| 5 to under 6.5 | 5.0 Fair |
| 3.5 to under 5 | 3.5 Poor |
| below 3.5 | 2.0 Very poor |
Overall
The overall score is a weighted average. Safety counts most:
| Safety | 40% |
| Walkability | 30% |
| Public transport | 20% |
| Quiet | 10% |
When a score is missing, the others are averaged with their weights. Nightlife is not part of the overall score: more bars is good for some trips and bad for others. For the same reason its labels say how lively a place is (Very lively, Lively, Moderate, Limited, Very little) rather than how good, and it has one color.
Safety
Safety is rated for the neighborhood, not the exact building. An AI research assistant searches the web, in English and in the local language, for what is reported about the area: local news on crime, police and city statements, travel advice that names it. It rates safety by day and at night from 1 to 10 on one scale for every city: 10 means walking alone at any hour is unremarkable, 6 typical big-city caution, 2 that locals and official advice say to avoid it. A city's neighborhoods are then rated side by side, so that where one is safer than another, the ratings show it.
The research starts from official figures for the whole country and city: the homicide rate (UNODC), the US and UK governments' travel advice, and in European cities the share of residents who feel safe walking alone at night (Eurostat). They set the level of the city; what is known about the neighborhood moves it up or down from there.
In England, Wales and Northern Ireland, and in New York, Los Angeles, Chicago, San Francisco and Philadelphia, the report also shows police-recorded crime within 1 km of the neighborhood over the last twelve published months (data.police.uk in the UK, each city's open-data portal in the US), by kind of crime and compared with the median of the city's neighborhoods on this site. US cities record crime in their own categories, so compare a neighborhood with others in the same city, not with another city. These figures are not part of the safety score. Police counts aren't adjusted for how many people are there: busy central streets record many times the crime of residential ones because far more people pass through, so a high count there doesn't mean a higher risk to one visitor. We tested using them in the rating, and it rated London's busiest central areas below neighborhoods that locals and police reporting single out.
The safety score is 40% the daytime rating and 60% the night rating, because the night is when most visitors worry. Bars, street lighting or tourist sights on the map play no part, and neither do the residents' ethnicity, religion, nationality or income. Where the research found nothing about the safety of the area or its city beyond the country figures, the score says “Limited data”. The pages it used are listed under the safety section of each report.
Walkability
Walkability is calculated for the exact address from the places around it. For each everyday need we find the nearest place within 800 m:
| Groceries (supermarket, bakery, corner shop…) | 25% |
| A place to eat | 20% |
| Health (pharmacy, doctor, clinic) | 15% |
| Café | 10% |
| Services (bank, ATM, post office) | 10% |
| Park or garden | 10% |
| Choice of food: 10 or more places to eat or drink within 500 m | 10% |
A need counts fully when the nearest place is within 300 m, and less the further it is, down to a quarter at 800 m. A need with nothing within 800 m counts nothing.
Public transport
Public transport is calculated for the exact address from the stops within 800 m:
- 40% how close the nearest stop is (full marks next door, nothing at 800 m),
- 40% how many stops there are, with stops within 400 m counting three times as much as those further away,
- 20% how many kinds of transport stop nearby: bus, tram, metro, train or ferry.
Quiet
Quiet is rated for the neighborhood, from the same research as safety: how noisy the area is by day and at night, from 1 (very quiet residential) to 5 (very loud: a major road, a nightlife hub, an airport approach). Doubled to 0 to 10, the quiet score is 10 minus the noise rating, with the night counting 60% and the day 40%.
Once a report's night & noise section is in, what is mapped right next to the address counts too: motorways and main roads, railways and tram lines within a few hundred metres, a large airport's runway or approach path within 10 km, and nightclubs or bars open past midnight on the same street. Each counts most when close and nothing past its reach (a main road from 250 m, a railway from 400 m); the loudest counts in full and the next ones less. Together they take up to 4 points off, scaled by how quiet the neighborhood is, because a noisy neighborhood's rating already includes much of its traffic.
Nightlife
Nightlife counts the bars, pubs and clubs within walking distance, including cafés that are really café-bars (as in much of the Adriatic and the Balkans). Each one adds a point, up to 10.
Where the data comes from
- Places and stops: OpenStreetMap, the map anyone can edit, through Geoapify.
- Roads, railways, runways, late opening hours and street lighting: OpenStreetMap, through the Overpass API. Opening hours and lighting count only where they are mapped.
- Safety, noise and what the area is like: AI research of public web pages, with links to the ones it used, starting from official figures: UNODC homicide rates (through Our World in Data), US State Department and UK Foreign Office travel advice, Eurostat's survey of how safe residents feel, and, shown alongside but not scored, police-recorded crime in the UK from data.police.uk (Open Government Licence) and in five US cities from their open-data portals (NYC Open Data, Los Angeles Open Data, the Chicago Data Portal, DataSF, OpenDataPhilly).
- The summary: written by AI from plain-language facts taken from the scores ("very walkable", "noisy"), never from the numbers themselves, so it can't contradict them.
- Live ratings and opening hours (paid plans): Google Maps, loaded in your browser when you ask. They are not part of any score and are never stored.
- Locations: Google, OpenStreetMap and the listing itself. Airbnb shows only an approximate location, so those reports describe the area within about 500 m.
How fresh it is
Places, stops and the neighborhood research are refreshed after 90 days, the next time someone opens the report.
What the scores can't tell you
Maps miss places and research can be out of date. A single street can be louder or calmer than its neighborhood. Use the scores to compare places and to know what to ask, and check the details that matter to you. If something in a report is wrong, open the report and use Report a problem in its menu.