FAQ

WORK IN PROGRESS. This census is an alpha, not a finished dataset.

How finished is this?

Not very. This is a work in progress, more an alpha than a beta. Thousands of candidate records are still under human review, and grades, counts, and positions will change between builds. Treat every number here as provisional and check the build stamp in the footer before citing anything.

What counts as a facility?

It holds people in custody against their will, overnight or longer. Jails, prisons, detention centers, juvenile detention, work camps, custody reentry, forensic psychiatric commitment, immigration detention, tribal jails, military brigs. Not offices, courts, probation, civil commitment, closed facilities, or museums.

Is every dot a confirmed facility?

No. The default view is the published census only. Everything else (candidates, unverified leads, records held back) is opt-in, drawn differently, and labelled as such. A name-inferred record is a candidate, never “confirmed.”

Why are some records shown at the county center?

Some records were discovered through the Google Places API, and for a subset of them that search is still the only source attesting their precise location. Google’s terms restrict republishing its data, so this site does not serve those coordinates. Instead, those records appear as small dots at national zoom levels (drawn into the map images, not served as data) and as hollow county-center markers when you zoom in. Once a record is independently confirmed by a roster, a quoted official page, or the federal HIFLD map, its true position appears. Positions that are shown come from public sources or are corroborated by them.

What are the outlines and hollow markers?

Outlines are HIFLD polygons matched to a record. Hollow markers sit at a city or county centroid, not at the facility itself; the data says so rather than implying precision it does not have. Records with no coordinates are absent from the map, never placed falsely.

Does the census count people?

Never. Facilities only. Sources with individual custody records were excluded at retrieval, not scrubbed afterwards. Population figures are aggregates published by the operating agency.

What are the aerial photos?

USDA NAIP imagery (public domain, 60 cm), with recorded acquisition years. They are there so a person can look; no claim in the census rests on what software saw in a photograph. Click any photo to view it full screen.

Where do the numbers come from?

Computed fresh from the dataset build stamped in the footer, by the same script that generates the map data. Nothing is hand-typed. The census is a work in progress; the published layer grows as evidence admits records.

What is this built with?

Hugo, Leaflet, a Python extraction script, docker compose. Fully static. The only network request is the optional OpenStreetMap basemap.

Sources and licensing

Rosters and documents come from government and operator publications. Geocoding uses US Census TIGER/Line street files (public domain). Aerial chips are USDA NAIP (public domain). The comparison layer is HIFLD Prison Boundaries, public federal data with closed facilities excluded. Discovery used the Google Places API, and one overlay comes from an SEO aggregator of jail listings; both are corroboration, never primary evidence. The basemap, when enabled, is © OpenStreetMap contributors.