Built from the June 2026 National Address Register release. What the register is

Data dictionary

Every field in the address, unit and postal code downloads, what it means, and where it comes from.

Source release: Statistics Canada National Address Register, June 2026 (Catalogue 46-26-0002) Build date: [fill] Coverage: Canada, by province, census division and census subdivision Formats: GeoPackage (.gpkg), CSV Published projection: EPSG:4326 (WGS 84) Boundary vintage for CD/CSD attribution: [fill] H3 resolution: [fill]

Three files, three grains

FileOne row perUse it for
LocationsPhysical buildingMapping, joins to parcels or boundaries, dwelling counts
AddressesBuilding unitUnit-level detail, joined to the location file
Postal codesPostal codeDensity, catchment and market analysis

The location and address files follow the source data model directly. Statistics Canada defines a location as a physical building and an address as a building unit, so a detached house is one location with one address, and an apartment building is one location with many. Joining the two files on location_id reconstructs the source structure exactly.

No source values are altered, corrected, interpolated or substituted. Coordinates are reprojected without adjustment.

Locations

One row per physical building. This is the file most users want.

FieldTypeSourceNotes
location_idtextlocationIdJoin key to the address file
civic_numbertextcivicNumber.number
civic_number_suffixtextcivicNumber.suffix
street_nametextstreetName
street_typetextstreetType.code
street_directiontextstreetDirection.code
city_nametextcityName.en
province_codetextprovince.codeNumeric province code
building_usageintegerbuilding.usageSee code table below
src_xnumericbuilding.dominionLandSurvey.xSource coordinate, untransformed
src_ynumericbuilding.dominionLandSurvey.ySource coordinate, untransformed
latitudenumericReprojected from source coordinate
longitudenumericReprojected from source coordinate
postal_codetextmailingAddress.postalCodeNull where units disagree
location_unit_countintegerDerived
cduidtextDerived
csduidtextDerived
h3_[res]textDerived

Building usage codes: 1 residential, 2 partial residential, 3 non-residential, 4 unknown.

The unit number lives in the source civicNumber.prefix field. It is not published here, because a location-grain row does not represent a specific unit. It is in the address file.

Coordinates

The register carries two positions. We publish the building coordinate, reprojected to WGS 84. The register’s other coordinate, derived from a blockface centroid — a street-segment midpoint rather than a building — is not published.

src_x and src_y are the untransformed source values, in the projection used by the register. They are included so the reprojection can be checked. In the GeoPackage the point geometry and the latitude and longitude columns hold the same values; they are not two independent positions.

Positional quality varies with the contributing source and is not consistent between them. We publish the source coordinate at face value.

postal_code

The postal code for this location, where all units share one. Where units carry different postal codes — common in larger buildings — this field is null and the address file carries the detail.

location_unit_count

The number of source address records at this location. A value of 1 is a single unit, typically a detached dwelling. Greater than 1 is multiple units in one building.

Equals the number of rows for this location_id in the address file. A dwelling-count proxy at building level, which most municipal open address files cannot produce because rental units are frequently absent from them.

cduid and csduid

Census division and census subdivision identifiers, assigned by spatial attribution against published boundaries.

The source carries a census subdivision code at 2016 vintage, and no census division code at all. These fields supply both at the vintage stated above, so the data joins to current census products without a spatial operation.

  • These are our assignments, not Statistics Canada’s.
  • Where a source coordinate is wrong, the address is attributed to the wrong subdivision. Counts aggregated by census subdivision are approximate and are not authoritative municipal totals.
  • Records without a coordinate carry no value in these fields.

Check the boundary vintage before joining. Subdivision identifiers change between census cycles through amalgamation, boundary revision and new incorporation. Joining to a different cycle’s table produces silent mismatches.

h3

Uber H3 cell index for the location coordinate, at the resolution stated above. A spatial index, not an identifier — many locations share a cell. Provided so density aggregation and hex-level summaries can be done with a grouped query rather than a spatial operation. Coarser resolutions can be derived from the value supplied. Null where there is no coordinate.

Addresses

One row per building unit. A link table, not a standalone dataset: it carries only what varies between units at a location. Street address, city, coordinates and everything else are location-level and live in the location file. Join on location_id.

FieldTypeSource
address_idtextaddressId
location_idtextlocationId
unittextcivicNumber.prefix
postal_codetextmailingAddress.postalCode

Use this file where unit-level detail is required — postal codes for individual units, or any work needing one row per dwelling. The number of rows for a given location_id equals location_unit_count in the location file.

Postal codes

One row per postal code, with a derived point.

FieldTypeSource
postal_codetextmailingAddress.postalCode
postal_code_unit_countintegerDerived
latitudenumericDerived
longitudenumericDerived
cduidtextDerived
csduidtextDerived
h3_[res]textDerived

Units are counted under their own postal code only. A building spanning several postal codes contributes its units to each according to where they fall, not its total to all of them.

Positions are calculated from register addresses. They are not certified for mailing, and are not affiliated with or endorsed by Canada Post.

Coverage is incomplete and compounds the gaps described below. A postal code absent from this file may exist — absence is not evidence of non-existence.

Known limitations of the source

These are properties of the register, not of our processing. We do not correct them.

Addresses are incomplete. An address is published only once validated against at least two independent sources. Combined with the semi-annual release cycle, coverage is systematically weakest in recently built areas — new subdivisions are precisely where two independent sources have not yet agreed. If your work concerns growth areas, treat the register as conservative.

Positional quality is inconsistent. Coordinates originate from different contributing sources and vary in quality between them. We publish them as supplied.

Not every record has a coordinate. Those records cannot be attributed to a census division or subdivision and carry no H3 value.

What these files are not

  • Not corrected. Positions are as published by Statistics Canada, reprojected without adjustment.
  • Not a parcel or property dataset. No boundaries, ownership or assessment information.
  • Not a mailing list. No occupancy information is present at any address.
  • Not current. The source is published twice yearly. The build date above states currency.

Format notes

GeoPackage carries null geometry for records without a coordinate and preserves full field names.

CSV has no null. Missing values are empty strings. Provincial files exceed the row limit of common spreadsheet applications — use the census division files if that is your tool.

Both formats carry identical fields and identical values.

Source and licence

Adapted from Statistics Canada, National Address Register, Catalogue 46-26-0002, June 2026 reference period. This does not constitute an endorsement by Statistics Canada of this product.

Use of the source publication is governed by the Statistics Canada Open Licence Agreement.