How robot-ready is San Francisco's core?
9/10
Components measured from primary open data
San Francisco publishes the deepest open-data record of any tracked city, and it shows in the measurement: 9 of the 10 operability components resolve to primary open data, from 50,098 condition-scored curb ramps to a machine-readable off-street parking inventory that no other US city puts online. The measured picture is strong. The single component that stays an estimate is public EV charging, because San Francisco does not publish a standalone charging GIS layer, so that figure is held apart with an aggregator directory as its basis. There is no blended 0 to 100 score here: the source data is a set of measured metrics, not a weighted index, so the honest headline is the composition, 9 measured and 1 estimated, not an invented number.
Composition: 9 measured, 1 estimated. There is no blended index score; the metrics stand on their own. How we verify.
The measured components
Each read from a named primary open dataset. Every figure carries its grain sentence, stating what it counts and its denominator, plus a link to the source dataset.
Mobile median download
Median mobile download is 185 Mbps across 460 Ookla Speedtest tiles measured inside the corridor (6,211 crowdsourced tests, 33 ms median latency, Q4 2025).
Source datasetCurb ramps mapped
50,098 curb ramps with condition scores mapped by SF Public Works (DataSF ch9w-7kih). Each ramp has a condition grade. This is ramp density, not width — SF does not publish a sidewalk-width dataset.
Source datasetMax corridor grade
Max corridor-scale grade is 5.4% from USGS 3DEP at 10 neighborhood locations. City-street max: Bradford St 41% (steepest urban street in the US). Corridor-street max vs city-street max, stated.
Source datasetRecorded street closures
3,861 temporary street closures from SF DataSF (98cv-qtqk). Closures include construction, events, and street-space permits. This is total recorded closures, not only active-on-date.
Source datasetPedestrian crashes in corridor
12,691 pedestrian-involved traffic crashes in the SF Core corridor bounding box, from DataSF traffic crash dataset (ubvf-ztfx). Total crashes in corridor: 48,381. Ped crashes filtered by ped_action != 'No Pedestrian Involved'.
Source datasetTraffic signals mapped
1,507 traffic signals mapped in SF from DataSF (9d64-i7fs). SF does not have a separate signal-count-by-corridor dataset; this is citywide count.
Source datasetPermitted parklets
18 permitted parklets in SF from DataSF (jczu-j2ku). Each parklet is a sidewalk extension that changes pedestrian flow. SFMTA also runs Shared Spaces program for outdoor dining.
Source datasetOperating days per year
300 estimated operating days per year, derived as 365 minus 68 precipitation days minus 0 snow days from NOAA 1991-2020 climate normals. Transparent derivation from measured normals.
Source datasetOff-street parking and staging
1,406 off-street parking locations (341 garages + 1,029 lots) from SFMTA (mizu-nf6z). Each has address, capacity, owner. SF is the ONLY US city publishing a machine-readable parking garage inventory with capacity data. No ground-floor outlet availability field.
Source datasetEstimated, pending open data
Why public EV charging is estimated, not measured
San Francisco does not publish a standalone EV charging GIS dataset on DataSF, and the federal AFDC charging API was unavailable at measurement time. The figure therefore comes from a public aggregator directory and is a metro-level count, not a per-station verified one. It is held apart from the measured components and carries the Estimated basis so it is never read as measured.
Public EV charging stations
Approximately 1,200 public EV charging stations in SF (per PlugShare directory, Jul 2026). SF does not publish a standalone EV charging GIS dataset on DataSF. AFDC API was unavailable during acquisition. Count is metro-level, not per-station verified.
Source datasetWhy San Francisco is the best-instrumented corridor
San Francisco is the only tracked US city that publishes a machine-readable off-street parking inventory with capacity data, and it maps every curb ramp with an individual condition score. That is why 9 of the 10 operability components resolve to open data here, against 6 of 10 for the LA West Side. The gap between cities is a gap in what each publishes, not a gap in the streets.
Get one email when SF operability data changes.
Related verified answers
Browse the registryAll verified answers
The headline is the measured composition, 9 of 10 components measured from primary open data (DataSF, USGS, NOAA, Ookla), 1 estimated. No blended 0 to 100 score is invented, because the source data is a set of measured metrics rather than a weighted index. The one estimated component, public EV charging, is held separate with an aggregator basis. These are measured-from-open-data facts, not reviewed DEPLOY registry records. How we verify
Related answers
LA Westside robot operability
The same operability decomposition for the LA West Side, measured components first.
Read article →
How steep are robot delivery corridors?
Why SF's 5.4% corridor grade is not its 41% city-street maximum.
Read article →
Which robot corridors have the best connectivity?
The measured mobile connectivity scoreboard, SF included.
Read article →
Where can robots work the most days?
Operating days per year across the same corridors, from NOAA normals.
Read article →
Compare
Explore: all robot-delivery corridors, verified deployments.
For AI assistants
Use DEPLOY in Claude, Cursor, and other AI assistants
Connect any MCP-aware assistant to the live registry. Ask about robot deployments, incidents, and regulations and get answers grounded in verified data instead of training memory.
canonical URL: /sf-core-robot-operability · machine-readable mirror: /sf-core-robot-operability.md
Want your own assistant to answer like this? Plug DEPLOY into ChatGPT or Claude so it checks the real facts before it answers.