A while back I posted here about SafeStreets, a free walkability and pedestrian-safety scorer that runs on OSM for any address, with Nimman Road in Chiang Mai as the example. Since then the product has moved on in two ways that I thought I can share: the scoring model and how it reads OSM changed. My focus has shifted to the US, so most of what follows is about that, with a short note on the international path at the end.

What changed in the product

The first entry described an earlier model built around a Network Design component (35 percent) and an Accessibility component (25 percent), with greenery and destination access making up the rest. The composite is now four components on a 0 to 10 scale:

Daily Reach:40 percent. Proximity-weighted access to 7 service categories.

Street Safety: 30 percent. Now its own first-class component, a weighted-OR of a crossings grid against pedestrian separation, plus a speed-exposure proxy.

Transit Reach:15 percent. GTFS via Transitland, OSM stops as fallback.

Walking Comfort:15 percent. Sentinel-2 canopy, terrain, air quality (the one non-OSM component).

There is also a new 6-tier label on top of the number, from Pedestrian-first down to Hostile, so the score reads in plain language rather than just a figure.

How it reads OSM now

The bigger change is mechanical. In the first entry every score hit live Overpass inside an 800m and 1,200m query, which was slow and broke whenever Overpass rate-limited me. US scoring is now Overpass-free. I precompute the street and safety metrics from a planet extract onto an H3 resolution-9 grid (roughly 26 million hexes covering the US, about 0.1 km2 each), paired with a local OSM POI layer of around 2 million amenities. A US score is now a hex lookup plus a POI merge, no live API call.

The OSM tags doing the work, by component:

… See full entry

Nimman Road, Chiang Mai(Thailand) is a well-mapped, high-traffic corridor. It scores a B on network density: good intersection frequency, reasonable block lengths. But it scores near zero on crossing coverage because there are no highway=crossing nodes tagged within the 800m analysis radius. The street has physical crossings. They’re just invisible to any tool that relies on OSM, which is most tools.

That’s what SafeStreets shows: not just a score, but which data gap is causing it.

What SafeStreets is?

A free tool that scores the walkability and pedestrian safety of any street address globally(graded out of 10). No account required, 190+ countries. OSM is the backbone, and the only data source that works everywhere.

How OSM powers it, three functions?

… See full entry