Walkably1Walking to work, neighbourhood by neighbourhood
In France, in England and Wales, in Ireland, in Scotland and in Northern Ireland, the census counts how people get to work in each small neighbourhood. In every city, county or council area, we put the neighbourhoods in order twice, once by the share of people who walk to work and once by our score. Then we measured how well the two orders agree. That's a number from −1 to 1 (statisticians call it a rank correlation). At 1, the two orders are the same. At 0, there's no link at all. We set the pass mark at 0.35 before we saw a single result.
Francemiddle city of 28
Northern Irelandmiddle of 11 districts
Irelandmiddle of 34 counties and cities
England and Walesmiddle of 301 council areas
Scotlandmiddle of 32 council areas
pass mark, set before any result
We chose the formula on France. Then we ran England and Wales without changing anything, and the score passed there too. Walking to work is one kind of walking, which is why the pass mark is 0.35. In Paris, almost everything is a few minutes from everywhere. In the one Paris district in the check, the 16th arrondissement, the match is only 0.17.
Then we ran the same check, with nothing changed, in Scotland (census 2022) and Northern Ireland (census 2021). Both pass the 0.35 mark. In Scotland, the middle of 32 council areas agrees at 0.46. In Northern Ireland, the middle of 11 districts agrees at 0.70. There, a score built only on how close the daily needs are (we call that part Reach) agrees just as well (0.70), so the street layers don't raise the result.
The 2021 census in England and Wales took place during a COVID lockdown, when many people worked from home.
We also ranked the same neighbourhoods by how many people live there per km², the way a population map does. Our score agrees with walking to work much better than that ranking, in both countries where we ran it: France 0.73 against 0.52, England and Wales 0.63 against 0.41. It did better in all 28 French cities and in 288 of the 301 towns and cities in England and Wales. This was a separate run, so England and Wales comes out at 0.63 here, where Figure 1 shows 0.62.
Does the score only follow how rich a neighbourhood is? In Paris it doesn't: across 864 neighbourhoods, the score and income barely agree (0.07). In Chicago they agree strongly (0.66 across 365 census tracts), more than a population map does there (0.38). That's one city on each side, so it's only a first look.
How we place map cells in neighbourhoods moves these results by 0.002 at most (France 0.73, Ireland 0.69), so both still pass.
In Ireland, a small neighbourhood that sits inside a bigger one can end up with no map cells. In France, a cell on the line between two neighbourhoods counts for only one of them. A stricter way of placing the cells changes neither result by more than 0.002.
2A weaker check, between towns, not between neighbourhoods
In Czechia, Portugal and Italy, we also ran a check between towns instead of between neighbourhoods, and it's weaker. In each district, we put its towns or parishes in order by walking to work and by our score. Between towns, size drives both numbers: a bigger town has more shops and more jobs close to home. So this check is easier to pass, and it isn't part of Figure 1. The pass mark is the same 0.35, set before any result.
Czechia: passes (0.44, pass mark 0.35)
Across 63 districts, the middle district agrees at 0.44, but our street layers add nothing here: Reach alone gives 0.45. Inside Prague, across its 57 districts, the score agrees at 0.71, and Reach alone at 0.74.
Portugal: fails (0.13, pass mark 0.35)
Across 20 regions, the middle region agrees at only 0.13, and Reach alone gives 0.13 too. A likely reason, which we haven't tested: in Portugal, fewer people walk to work in bigger parishes, while our score goes up with parish size.
Italy: fails (−0.03, pass mark 0.35)
Italy's 2011 census (ISTAT) counts how people get to work in each comune (town). Rome and Milan are one comune each, so Italy gives no check between neighbourhoods, only this weaker one between towns. In each of 65 provinces, we put its towns in order by walking to work and by our score. The middle province agrees at −0.03: no link. 29 of 65 provinces are above zero, and 5 reach the mark (Milan is the highest at 0.58). In several provinces of the northern plain, such as Padua and Bologna, the order runs the other way. We haven't tested why.
3Two more checks inside cities
Here are two more checks that compare our score with other data inside one city each: Amsterdam and Madrid.
Amsterdam, a 15-minute walking index: agrees closely (0.74, pass mark 0.6)
In 2026, researchers published a "15-minute Amsterdam" walking index (Koljensic, figshare, CC BY 4.0; analysis year 2022). It uses a different method: walking areas drawn by a route planner (OpenRouteService), 15 kinds of place and the CBS population grid. Across 6,910 populated blocks of 100 m, our scores put places in nearly the same order: 0.74 (0.80 across 449 squares of 500 m). We set the pass mark, 0.6, before we joined the two. Both use places from OpenStreetMap, so this shows the method is sound, not that the map is complete. The census checks and the place checks cover that. The two differ most in a few single blocks, where our route rules cut the walk short (we are fixing this), and in Noord, where their index also counts metro and tram stops and leisure places that we don't.
Madrid, a travel survey of all trips: agrees (0.44, pass mark 0.35)
In 2018, Madrid's regional transport authority (CRTM) asked households about every trip they made (Encuesta Domiciliaria de Movilidad). It is a travel survey of all trips by where they start. It is not a census, and it does not cover only trips to work. Across 530 zones of Madrid city, we put the zones in order by the share of trips made on foot and by our score. They agree at 0.44 (likely between 0.36 and 0.51), above the pass mark of 0.35 that we set before the run. This is a moderate agreement, not a close one. Reach alone gives 0.41, so the street layers add a little. The whole Madrid region, which includes the city, gives 0.43.
4A test outside Europe
Ten US cities: passes narrowly (0.36, pass mark 0.35)
Our map doesn't cover the United States, so we ran the same score in the 10 largest US cities, only to test the method somewhere new. The Census Bureau's American Community Survey (2020 to 2024) counts how many people walk to work in each census tract, a small neighbourhood. We compared about 3,000 tracts within 10 km of each city centre, the same way as in Europe. The middle city agrees at 0.36, just over the pass mark of 0.35, which we set before any result. So this check passes, narrowly. Our first run on 10 Oct gave 0.34: it read speed limits in km/h as mph, used Spain's sun clock and had no US tree data. Philadelphia agrees best (0.60) and San Antonio least (0.09). In cities like Houston, Phoenix and San Antonio, about 1 person in 100 walks to work, so there's little to rank. New York is low for the opposite reason: most of the centre is walkable, and people go to work by subway. A few map cells sit on the border between two census neighbourhoods, and the order we match them decides which neighbourhood gets each cell. A stricter way of matching changes the US figure by at most 0.002, so the check still passes.
In Chicago, where a travel survey records every trip, the score tracks how much residents walk for all their daily trips (0.53) much more closely than walking to work (0.26). Source: CMAP, My Daily Travel 2018-19.
5Are the places on the map really there?
We drew 291 addresses in 31 countries and checked 80 of our answers one by one against other sources. When we name a place near you, it was there about 91 times in 100. When we say nothing is within 20 minutes on foot, that held about 85 times in 100; most misses were places missing from OpenStreetMap. We repeat this check with each data release.
We checked 600 homes in England, drawn at random, more often where more people live. We compared our map with two official lists: the Department for Education's list of schools and the Department for Transport's list of public transport stops. For about 3 to 5% of homes, one of these lists put a school or stop 3 or more minutes closer than our map.
6The real walk, not a straight line
Walkably measures the real walk on real streets. The straight-line shortcut many maps use is off by 3 minutes or more about 1 time in 9, and almost always makes places look closer than they are. That's what our first test found, in 6 countries; a test in every country replaces these figures after the next data release, and we repeat it with every release.
On 27 Oct 2026, we drew 67 places at random in Germany, Estonia, France, the United Kingdom, Italy and the Netherlands: 17 in big city centres, 20 in smaller cities, 16 in suburbs, 10 in small towns and 4 on islands. For each place and each of the eight daily needs, we took the nearest one in a straight line and turned that distance into minutes the way the shortcut does: the straight line times 1.3, at 80 metres a minute. Then we compared that with the walk on the streets of OpenStreetMap, at the same speed.
Of 513 answers, 56 were off by 3 minutes or more (10.9%; the real share is probably between 6% and 16%). In 51 of those 56, the shortcut said the place was closer than it is. Half of all answers are off by half a minute or less, so the risk is in the big errors. Suburbs and islands do worst, about 1 answer in 6, but they are only 20 of the 67 places.
We sent the 56 big errors again to a public walking route planner (routing.openstreetmap.de), and 54 gave the same result. One was our mistake: a park in Reggio Emilia is about 200 metres away, and our route said 626 metres. We're finding the cause. Without that answer, the share is 10.7%. Our walks are routes on the map, not GPS records of real trips. Download every answer (CSV), the 67 places (CSV), the route planner check (JSON) and the script that counts them again (Python).
7Checks that didn't pass
Every pass mark was written down before we saw any result. We report the checks that failed as they are.
City against city: fails (0.00, pass mark 0.50); walk or public transport: passes (0.60)
Eurostat publishes the share of people who walk to work for whole cities. Across 195 cities in the first 20 countries on the map, our city scores don't agree with it at all (0.00, against a pass mark of 0.50). The measure has a flaw of its own. Where a metro carries people to work, few of them walk the whole way, so greater Paris (7.8%) and Lille (7.1%) rank below Clermont-Ferrand (14.5%), a city with no metro. So before we ran the check again, we wrote down a second version with the same pass mark: the share of people who walk or take public transport to work. It agrees much better, at 0.60, and passes the mark of 0.50. We report both results and don't tune the score to them.
Summer shade: fails (−0.18, pass mark 0.40), so the map shows it as an estimate
A study of 59 European city centres (see Credits) measured how much of each street has trees over it. In the 43 centres we compared, our shade doesn't agree with it (−0.18, against a pass mark of 0.40). The audit counts trees only, while we also count the shadows of buildings. So before the next run, we wrote down a second check: our shade from trees alone. It agrees better, at 0.27, but it's still under 0.40. The two also count in different ways. The audit counts a whole street as shaded when trees cover more than a quarter of it. We count only the part of the street that is in shadow. In central Barcelona, the audit finds 95% and our trees alone give 26%. So the map shows summer shade as an estimate and says so. The same study did confirm our busy-road measure, at 0.56 across the same 43 centres, above its pass mark of 0.40.
Barcelona, street by street: fails (−0.03, pass mark 0.40)
Alexandros Bartzokas-Tsiompras, one of the authors of the study of 59 European city centres (see Credits), shared his own street-by-street ratings for the Eixample in Barcelona with us. Trained observers rated close to 2,000 street segments from street images on 14 items, and the images were taken in 2018 and 2019 (Data in Brief, 2021; Journal of Transport Geography, 2023). The ratings cover the Eixample only, and some streets changed since then. Our street quality doesn't agree with the audit's overall score (−0.03, against a pass mark of 0.40). Many sidewalks and bike paths in Barcelona carry the street's name on the map, so some segments were matched to a sidewalk, not the road. So before we ran it again, we wrote down a second version with the same pass mark: match each segment to the road only. It gave −0.07, so matching to sidewalks was not the cause. Our shade from trees alone agrees at 0.10, against a pass mark of 0.40: the right way, but far too weak, so it fails too. We publish the result as it is. When we looked for the reason afterwards, we found two. Most of the audit's 14 items rate things our score doesn't measure, such as parks, bike lanes, benches and graffiti. And our shade estimate misses many street trees. We found these after the check failed, so they don't change its result.
Against a simple count of places nearby: fails (France 0.73 against 0.78)
Before the run, we wrote down that our score should beat two simple rankings by 0.05 or more, in France and in England and Wales. One ranks neighbourhoods by people per km². The other is a plain count of shops, schools and other places within 800 m. Our score beats the ranking by people but not the count of places: France 0.73 against 0.78, England and Wales 0.63 against 0.67. The street-quality half of the score, on its own, beats neither. We keep the score because it says how far each daily need is and how good the streets on the way are, which a count can't. We report the result as it is.