What does the score mean?

The a11y-Score provides a score between 0 and 100. A higher value means better accessibility. A value of 100 means that all included accessibility criteria are fully met.

Not a percentage

The value is not a percentage! A score of 50 does not mean that “half of a region is accessible” or that accessibility is “half as good” as with a score of 100. It also does not mean that only half of all criteria are met.

Instead, the value should be understood as a continuous scale. You can think of it like this: if people are asked to rate how satisfied they are on a scale of 1 to 10, a person with a value of 6 is not necessarily twice as satisfied as a person with a value of 3. But they are definitely more satisfied.

How reliable is a score?

A score is an average over all places that have data. As with any survey, the fewer answers there are, the more the result depends on chance. With only a few places, a single new entry can shift the score a lot. With many places, one more entry hardly matters. Statisticians call this the law of large numbers.

An example from our data

Mobility scores for all places, as of November 2026:

• Traiskirchen (Lower Austria) 50 (5 Places with data): If one more fully wheelchair-accessible place is added, the score could jump to 58. If that place is not wheelchair accessible, it drops to 42.
• Berlin 60 (32,677 Places with data): One more place changes the score by about a thousandth of a point.

How much small amounts of data scatter becomes visible when comparing all 3,716 towns and municipalities in Germany and Austria that have a score. The middle 80 % of them lie in these ranges:

• 1 to 5 places with data: anywhere between 0 and 100
• 6 to 20 places: between 29 and 88
• 21 to 100 places: between 42 and 88
• 101 to 500 places: between 48 and 77
• more than 2,000 places: between 61 and 68

Small towns are not that much better or worse than large cities. Their scores scatter mainly because of chance: with few data points, every single one makes a big difference.

More criteria per place

The same applies to each individual place. If all that is known about a café is that its entrance has no steps, it scores 100 for mobility, even if its toilet might not be wheelchair accessible. With every additional criterion recorded, the value of a place gets more accurate and jumps around less. As long as only a few features are known, places often end up at the very top or the very bottom: in Berlin, at least a quarter of the places with mobility data score 0, and at least another quarter score 100.

How much data does a reliable score need?

A rule of thumb from statistics: uncertainty shrinks with the square root of the amount of data. To be twice as precise, you need four times as much data. For the a11y-Score, a rough and rather cautious estimate (a 95 % confidence interval) says how far the score could still be from the value it would have if every place of that kind were recorded:

• with data on 10 places: about 30 points up or down
• 25 places: about 20 points
• 100 places: about 10 points
• 400 places: about 5 points
• 10,000 places: about 1 point

From around 100 places with data, a score is reasonably reliable. Fine comparisons need a few hundred. Currently, 435 of the 3,716 towns and municipalities have reached 100 — for most of them, the score shows a rough trend at best.

Is a difference real or just chance?

Whether two scores really differ is a question of statistical significance: is the difference bigger than chance alone could explain?

• Hamburg and Berlin (67 with 16,781 places, 60 with 32,677 places): the uncertainty is below one point for each city, so the difference of 7 points is not down to chance.
• Traiskirchen and Melk (50 with 5 places, 63 with 4 places): here, chance alone can easily account for 40 points. These numbers cannot tell yet which of the two towns is more accessible.

And this is what it looks like today, live from our data: Hamburg 67 (16,781 places with data), Berlin 60 (32,677 places with data), Traiskirchen 50 (5 places with data), Melk 63 (4 places with data).

One limitation remains: this calculation assumes that the recorded places were picked at random. They are not — volunteers mostly record what catches their eye or what they need themselves. Such sampling bias cannot be calculated away. It only gets smaller as a larger share of a region's places is recorded.

What this means

Every new entry helps — not because it automatically improves the score, but because it makes the score more reliable. That's why every chart and every score on our page shows how many places a value is based on. Where there are only a few, the score is a first indication rather than a verdict.