Which Format to Take

The format every desktop GIS reads without being asked, and the one most county assessors still publish.

Take it when

Anything that has to open on someone else's machine without a conversation about drivers. Thirty years of tooling assumes it.

Watch out for

Field names are truncated to 10 characters and the DBF attribute table has no real text encoding declaration, so long assessor column names arrive mangled. The 2 GB per-file ceiling is a real limit on the biggest urban counties.

QGISArcGIS ProGeoPandasGDAL/OGRPostGIS shp2pgsql

Plain text, WGS84, readable in a browser and in a code review. The web's parcel format.

Take it when

Web maps, quick inspection, scripting, and diffing two vintages in version control.

Watch out for

It is uncompressed JSON, so a large county runs to hundreds of megabytes, and the spec pins coordinates to WGS84 — if you need the county's own state-plane CRS for area calculations, take the GeoPackage instead.

LeafletMapLibreQGISGDAL/OGRPython json

One SQLite file, full-length field names, real CRS metadata, no sidecar files to lose.

Take it when

Everything the Shapefile is bad at: long attribute names, UTF-8 owner names, files over 2 GB, and several layers in one artefact.

Watch out for

Older ArcMap installations and some legacy county workflows cannot read it. It is a binary database file, so it does not diff.

QGISArcGIS ProGDAL/OGRGeoPandasSQLite

The attribute table with the geometry left behind — parcel IDs, owners, addresses, valuations.

Take it when

Joining parcel attributes to something you already have keyed on APN, or loading assessment values into a spreadsheet or a warehouse without a spatial extension.

Watch out for

There are no shapes in it. If you need to draw or intersect anything, this is the wrong download.

ExcelpandasDuckDBPostgres COPY

Columnar, compressed, and typed. Built for reading one column of ten million parcels rather than one parcel at a time.

Take it when

Analytics at scale — multi-county studies, DuckDB, Spark, and anything where the file lives in object storage and you want to scan it without downloading all of it.

Watch out for

Desktop GIS support is recent and uneven; check your GDAL is 3.8 or newer before assuming QGIS will open it.

DuckDBGeoPandasApache SparkGDAL/OGR 3.8+