P-code format

P-codesPlace code: the unique identifier of an administrative unit, such as AFG.001.012. Every code starts with the code of the unit above it. follow a hierarchical structure built from three components:

Level examples:

The code is always hierarchical, any unit is a valid prefix of every code below it, enabling SQL prefix queries to return all children without a join.

P-codes carry no version information. Dataset version is tracked as metadata using version, valid_from, and valid_to fields. The system starts at version 1; previous COD-AB versions are linked through a migration table and migration services.

Country code

ISO3ISO 3166-1 alpha-3 three-letter country code (e.g. SYR = Syrian Arab Republic). is the standard for country codes, consistent with its use across statistical and humanitarian data systems.

For territories with contested or disputed status, P-codes use the ISO3 code assigned by UN geodata, including user-assigned ISO3 codes for areas outside standard country boundaries. Administrative units receive the ISO3 prefix for the territory they fall within, not the country administering them. Where a country administers both disputed and undisputed territory, its dataset spans two prefixes: the standard country code for undisputed units and the UN geodata user-assigned code for disputed units. User assigned prefixes always begin with a capital X.

Delimiter

The dot (.) is the delimiter. It is safe in SQL LIKE queries, safe in Excel (plain text, not a formula), safe in URLs, and self-documenting: AFG.001.012 reads as a hierarchy without explanation. The delimiter allows digit width to vary when needed. Because each level is separated explicitly, a code can be parsed into its components without relying on fixed positions.

Digit width

Digit width was determined empirically using a global boundary dataset (186,526 features, 285 countries/territories) and IBGE Brazil (5,572 municipalities). The relevant ceiling is not the global total but the maximum number of children any single parent has at each level.

Three digits clears every level across the full dataset. Systems using fewer digits produce incorrect string sort order in Excel, SQL, and every tool that sorts alphanumerically.

Minas Gerais ceiling watch case: Brazil’s Minas Gerais state has 853 municipalities as direct ADM2 children, occupying 85.4% of a 999-unit ceiling. Brazil has historically added municipalities at roughly 4 per major state per decade. A level that exceeds 999 units , A level that exceeds 999 units can therefore take a fourth digit (e.g. BRA.001.1000) without repadding or reissuing existing codes, previously assigned codes are not repadded: BRA.001.001 remains BRA.001.001, and the thousandth unit is assigned BRA.001.1000. This outcome is recorded in the country’s metadata.