Address Formats Around the World: A Developer’s Field Guide

Why there is no universal address format

Postal addresses were never designed; they accreted. Each country’s format grew out of its own history of land registration, urban planning, and postal logistics, and each national postal operator standardized whatever its mail carriers already understood. The Universal Postal Union coordinates delivery between countries, but it deliberately does not impose one layout — the address block is formatted by the destination country’s rules, not the sender’s.

For software this means there is no schema you can normalize everything into without losing information. A "state" field is mandatory in the United States, meaningless in the United Kingdom, and actively confusing in Japan, where the equivalent unit is a prefecture that appears at the opposite end of the address. Any single rigid schema is a bet that all your users live in one country. Generated multi-country addresses show you where that bet loses.

Number first or street first?

English-speaking countries generally put the house number before the street: 12 Main St in the US, 221B Baker Street in the UK, 4 Wattle Lane in Australia. Much of continental Europe reverses it: a German address reads Musterstraße 12, a Spanish one Calle Mayor 12, a Dutch one Hoofdstraat 12 — street name first, number last, often with suffixes like 12a or 12-14 attached to the number.

This matters most to parsing and display code. If you split a delivery line with a regex that expects leading digits, Musterstraße 12 either fails validation or gets its fields swapped. If you reassemble an address from structured components, hard-coding the order "number, street" prints German addresses backwards. Test both directions: feed number-first and street-first samples through every code path that splits, validates, or re-joins a street line.

Large-to-small versus small-to-large

Western addresses read small-to-large: recipient, street, city, region, country. Japan and China read the other way. A Japanese address starts with the postal code and prefecture, narrows to the city and ward, then to a numbered district, block, and building — often with no street name at all, because most Japanese streets are unnamed and buildings are numbered by block. A Chinese address written natively runs province, city, district, road, number, building, apartment, with the recipient’s name last.

The trap for software is assuming line one is always the most specific line. Sorting, deduplication, and "same address?" comparisons that key on the first line will treat two different Tokyo apartments as identical because both begin with the prefecture. Label printing is the other classic failure: a template that prints fields top-down in Western order produces an address a Japanese carrier has to read bottom-up.

What counts as a city?

Even the humble "city" field means different things in different countries. In the US it holds an incorporated municipality that pairs with a state. In the UK the conventional equivalent is the post town — a postal routing concept that may not match the town someone says they live in, and UK counties are optional in modern addresses. In Australia the field holds the suburb, a much smaller unit than an American city, paired with a state or territory abbreviation and a four-digit postcode.

The practical consequence is that "city" should be treated as a free-text locality field, not validated against a list of cities you happen to know. A dropdown of localities is only feasible per-country with authoritative data; anywhere else it becomes a wall that legitimate users cannot climb. Generated addresses from the US, UK, and Australia will each put a differently shaped value in this field — exactly the variety your validation needs to survive.

Postal codes come in many shapes

Postal codes are the field developers most love to over-validate. The US uses a five-digit ZIP such as 90210, optionally extended to ZIP+4. Germany and France also use five digits, but with different geographic logic and leading zeros that die in integer columns. Japan uses seven digits with a hyphen, NNN-NNNN, like 100-0001. China uses six digits. Australia uses four. Then come the alphanumerics: a UK postcode like SW1A 1AA has a variable-length outward code, and a Canadian code alternates letter-digit-letter, digit-letter-digit: A1A 1A1, conventionally with a space in the middle.

Three bugs to hunt for: the numeric-only assumption, where a number input or a \d+ regex rejects every UK and Canadian user; fixed-length checks, when valid postcodes run from four characters to eight or more with spaces and hyphens; and silent normalization that strips the space from SW1A 1AA or the hyphen from 100-0001 — perhaps fine internally, wrong on a shipping label. Generated addresses across these seven countries trip all three in a single test run.

Designing and testing address forms that survive contact

The safest design is fewer, looser fields: one or two free-form address lines, a locality line, an optional region, a free-text postal code, and a country selector that drives labeling. If you need structured fields, adapt them per country — relabel state to province or prefecture, drop it where it does not exist, and never apply one country’s postal code regex to another country’s users. Make fields generous: allow at least 100 characters per line, and store everything as Unicode end to end, because Musterstraße, 東京都, and 北京市 must survive the round trip through your database, APIs, and PDFs unmangled.

Then verify it like any other requirement: with test data. Generate addresses from a handful of deliberately dissimilar countries — the US, UK, Germany, Japan, China, Canada, Australia — and push them through every surface that touches an address: signup and checkout forms, server-side validation, database storage, search, deduplication, label and invoice rendering. Watch for truncated umlauts, rejected alphanumeric postcodes, lost leading zeros, and required fields with no valid answer. Each failure is a bug a real international user would eventually hit; generated data lets you meet it in staging instead of production.

Frequently Asked Questions