A Japanese address does not identify a building by a street name and a house number. Most of Japan has no street names at all. An address is a nested set of areas, narrowing from the prefecture down to a city block, followed by a three-part block number that locates the building inside that block.
This article describes the Japanese address format, how CDQ represents it in the CDQ address model, and which reference data CDQ maintains. It is the reference for anyone mapping Japanese business partner data into CDQ, or consuming curated Japanese addresses from CDQ.
In this article, you will understand:
- How a Japanese address is structured, and why it cannot be mapped onto a street-and-house-number model
- How each Japanese level maps to a concept in the CDQ address model, and which SAP field it reaches
- Why the block number belongs in one field and the house number stays empty
- Which reference data CDQ holds for Japan, and what the published values look like
- Where the current CDQ coverage for Japan is incomplete
Available in:
- CDQ Wiki: business partner metadata for Japan: browse the published Japanese metadata
Japanese addresses are area-based, not street-based. Instead of naming a line (a street) and a position along it (a house number), a Japanese address names a series of progressively smaller areas and then a position inside the smallest one. Outside a few historic exceptions, streets are simply not part of the address.
The consequence for an integration is direct: there is no value that means "street name", and the number that looks like a house number is not one.
Written largest-first, which is the order used in Japanese:
| Level | Japanese | Romanised | What it is | Example |
|---|---|---|---|---|
| 1 | 都 / 道 / 府 / 県 | -to / -do / -fu / -ken | prefecture | 東京都 (Tokyo) |
| 2 | 郡 | gun | rural district, groups towns and villages | only in rural addresses |
| 3 | 市 / 町 / 村 | -shi / -machi / -mura | municipality: city, town, village | 川崎市 (Kawasaki) |
| 4 | 区 | -ku | ward, a subdivision of a large city | 中原区 (Nakahara-ku) in Kawasaki |
| 5 | 町域 | cho-iki | neighbourhood | 丸の内 (Marunouchi) |
| 6 | 丁目 | chome | city block | 1丁目 |
| 7 | 番地 | banchi | sub-block or land parcel | 9番地 |
| 8 | 号 | go | building number within the sub-block | 2号 |
Below the address proper come the building, the floor and the room, which are premises rather than address levels.
A complete address, largest-first, with the postal code first as written in Japan:
〒100-0005
東京都千代田区丸の内1丁目9番2号
グラントウキョウサウスタワー 12FReading it: prefecture 東京都 (Tokyo), municipality 千代田区 (Chiyoda), neighbourhood 丸の内 (Marunouchi), block 1, sub-block 9, building 2, in the building グラントウキョウサウスタワー on floor 12.
The 丁目 / 番地 / 号 sequence is written three ways, and all three occur in business data:
| Notation | Example | Where it appears |
|---|---|---|
| Kanji | 1丁目9番2号 | official and postal use |
| Dash | 1-9-2 | most common in business master data |
| Mixed | 1丁目9-2 | both conventions in one value |
All three notations are equivalent. Dash notation is the most compact and the most widely used in master data, so it is the form CDQ treats as canonical for the block number.
In Tokyo's 23 special wards (特別区) the ward is the municipality: 千代田区 sits directly under 東京都, with no city in between.
In a designated city (政令指定都市) such as Kawasaki, Yokohama or Okayama, the ward is a subdivision of the city, so both levels are present: 神奈川県 (prefecture), 川崎市 (city), 中原区 (ward).
The same character 区 therefore means different things depending on which city it belongs to, and the level it occupies differs with it.
1-9-2 is block 1, sub-block 9, building 2. Only the final element, 号 (go), is even loosely comparable to a western house number. Splitting the sequence, or treating the first part as a street name and the last as a house number, breaks the meaning: the three numbers are only interpretable together, and matching on the whole sequence is far more reliable than matching on its parts.
Most addresses have no street name. There is nothing to put in a street field except the block number. An integration that requires a street name, or that promotes some other value into the street field to avoid leaving it empty, will corrupt Japanese addresses.
The prefecture suffix is part of the name. 東京 and 東京都 are not interchangeable: 東京 alone is ambiguous, since it can be read as the city rather than the prefecture. The same applies to 大阪 versus 大阪府. The suffix identifies the level and should not be stripped.
〒 is a symbol, not data. The postal mark prefixes a postal code for display. The stored value is the seven digits in nnn-nnnn form.
Kyoto does not follow the model. Kyoto uses traditional street-name references such as 四条通り上ル, describing a position relative to two crossing streets. These addresses have no chome-banchi-go sequence and cannot be decomposed into one.
Rural addresses use different area words. Where there is no chome numbering, an address uses 大字 (oaza) and 字 (aza) with a land parcel number instead. In these addresses 番地 refers to a land parcel rather than a sub-block.
Scripts mix freely. A single Japanese address can contain kanji, hiragana, katakana, fullwidth and halfwidth digits, and romaji, in any field. Building names in particular are frequently katakana transcriptions of English names.
When a Japanese address has been romanised or machine-translated upstream, several levels are often flattened into one value, for example Nakahara Ward Kawasaki City Kanagawa Prefecture. The component order is not stable, so the words Ward, City and Prefecture are the only reliable way to tell which part is which. Do not rely on position.
The CDQ address model has two parallel hierarchies. Administrative areas represent the political containers and support two levels, typed REGION and COUNTY. Localities represent the places themselves and support up to ten levels, typed CITY, DISTRICT, QUARTER or BLOCK. The neighbourhood and the block number are held together as a thoroughfare, and the building, floor and room as typed premises.
Japan's levels are placed as follows:
| Japanese level | CDQ concept | Type | SAP field |
|---|---|---|---|
| 都 / 道 / 府 / 県 prefecture | administrative area, level 1 | REGION | region |
| 郡 gun, rural district | administrative area, level 2 | COUNTY | county |
| 市 / 町 / 村 municipality, and a Tokyo special ward | locality, level 1 | CITY | city |
| 区 ward of a designated city | locality, level 2 | DISTRICT | district |
| 町域 neighbourhood, with the block number after it | thoroughfare value | STREET | street |
| 丁目 / 番地 / 号 block number | thoroughfare value, after the neighbourhood | STREET | street |
| house number | thoroughfare number | n/a | left empty, see below |
| building | premise | BUILDING | street supplement 1 |
| floor | premise | LEVEL | street supplement 2 |
| room | premise | ROOM | street supplement 3 |
This is the decision most likely to surprise someone reading a Japanese address out of CDQ.
Only 号 (go) resembles a western house number, and it is meaningless without the 丁目 and 番地 that precede it. Splitting only the last number out would:
- weaken matching, because
1-9-2is one semantic unit and comparing fragments of it produces false differences - break display, since reassembling the address then requires Japan-specific logic in every consumer
- push Japanese parsing downstream, into systems that have no reason to know about chome and banchi
So the whole 丁目 / 番地 / 号 sequence goes into the thoroughfare value as a single dash-delimited string, and the thoroughfare number is left empty.
The neighbourhood sits in front of it in the same field, giving 丸の内1-9-2 rather than a bare 1-9-2. That is how a Japanese address is written, the neighbourhood and the block number being one continuous unit, and it means the delivered street line names a place instead of consisting only of digits.
Where a receiving system insists on a populated house number, the sequence can be split as thoroughfare value 1-9 and thoroughfare number 2. This is a per-customer configuration rather than the standard, because the resulting values are less useful for matching.
| Rejected alternative | Why |
|---|---|
| Map the block number onto street name plus house number | There is no street name in a Japanese address, so the street field would hold a number and the house number would hold a fragment of it. Both values become uninterpretable on their own. |
| Split the sequence into three separate fields | No CDQ concept corresponds to 丁目 or 番地, and inventing three would make Japanese addresses incomparable with every other country. The sequence is also always written as a unit. |
| Hold the neighbourhood as a locality and leave the block number alone in the thoroughfare | This is the alternative that looks tidier and delivers worse addresses. A designated city has three values below the prefecture, the city, the ward and the neighbourhood, and only two locality levels are available, so giving one to the neighbourhood costs the ward its own field. It also leaves the street field holding nothing but 1-9-2, which is the single most common complaint about delivered Japanese addresses, and it forces the city field to become a compound string such as 川崎市中原区, which weakens name matching and duplicate detection. |
The address used above, expressed in the CDQ address model:
| Field | Value |
|---|---|
| Country | JP |
| Post code | 100-0005 |
Administrative area 1 (REGION) | 東京都 |
| Administrative area 1 short name | JP-13 |
Locality 1 (CITY) | 千代田区 |
Locality 2 (DISTRICT) | empty. 千代田区 is a Tokyo special ward, so it is the municipality and has no ward beneath it |
Thoroughfare (STREET) | 丸の内 1-9-2 |
| Thoroughfare number | empty |
Premise (BUILDING) | グラントウキョウサウスタワー |
Premise (LEVEL) | 12F |
The examples below show the shape of each dataset, not its full contents. The complete published lists for Japan are on the CDQ Wiki page business partner metadata for Japan, which also covers legal forms, identifier types and the integrated data sources.
All 47, with codes JP-01 to JP-47. The Japanese name carries its administrative suffix, so the suffix is part of the reference value rather than something to strip.
| Code | Japanese | Suffix | Latin |
|---|---|---|---|
JP-01 | 北海道 | 道 | Hokkaidō Prefecture |
JP-13 | 東京都 | 都 | Tokyo |
JP-14 | 神奈川県 | 県 | Kanagawa Prefecture |
JP-26 | 京都府 | 府 | Kyōto Prefecture |
JP-27 | 大阪府 | 府 | Ōsaka Prefecture |
A prefecture arrives written several ways in business data. All of the common forms are published as alternative names, so a correct value resolves however it is spelled. The canonical name keeps its suffix; a variant is an alternative to it, never a replacement for it.
| Variant kind | Example |
|---|---|
| Bare kanji, suffix dropped | 愛知 matches 愛知県, 大阪 matches 大阪府 |
| Romanised with the level marker | Aichi-ken, Tokyo-to, Osaka-fu |
| Romanised without the macron | Osaka matches Ōsaka, Hyogo matches Hyōgo, and the same for Hokkaido, Kyoto, Kochi and Oita |
| Uppercase | AICHI, OSAKA, the form SAP master data most often supplies |
| Level marker written out | Aichi Prefecture, with Prefecture recognised as the level marker in romanised text |
Six prefectures carry a macron in their Latin name: 北海道 Hokkaidō, 京都府 Kyōto, 大阪府 Ōsaka, 兵庫県 Hyōgo, 高知県 Kōchi and 大分県 Ōita. Each resolves with the macron and without it, because a source that cannot store the character is common.
A variant is added only when it can match exactly one prefecture. Japanese prefecture names are close enough in romanised form for this to matter: 岡山 Okayama beside 和歌山 Wakayama, 徳島 Tokushima beside 福島 Fukushima. A variant that could match either is worse than no variant, because the address it produces is wrong rather than merely incomplete.
Hokkaidō's 道 is integral to the name in a way that 都, 府 and 県 are not. There is no bare form 北海 and no Hokkaido-do. It carries its Latin spellings and nothing else, and its suffix should not be stripped.
| Reference data | Coverage for Japan |
|---|---|
| Prefectures | All 47, with ISO 3166-2 codes (JP-01 to JP-47) and names in Japanese and Latin script |
| Prefecture suffix | The Japanese name carries its administrative suffix, so 東京都 rather than 東京. One 道, one 都, two 府 and 43 県 |
| Prefecture name variants | Maintained. The bare kanji form, the romanisation carrying the level marker, the macron-free spelling and the uppercase form, as set out above. A prefecture written any of these ways resolves to the same record |
| Japanese address terms | Maintained. The level markers 都, 道, 府, 県, 郡, 市, 町, 村, 区, 大字 and 字; the block-number components 丁目, 番地 and 号; the Kyoto street word 通; and the premises vocabulary ビル, ビルディング, タワー, センター, プラザ, マンション, 階, 室 and 号室. Published in Japanese script, with romanised forms where these are written in Latin script |
| Rural districts (郡) | Accepted and preserved. Not yet validated against a CDQ list |
| Municipalities, wards, neighbourhoods | Accepted and preserved. Not yet validated against a CDQ list |
| Postal codes | Format only. Seven digits, accepted with or without the hyphen and canonically written nnn-nnnn. The postal code is not cross-checked against the area levels of the address, and Japan Post maintains the authoritative postal database |
The prefecture is standardised against reference data: the value is matched and the canonical Japanese name and ISO code are written back. Because the reference data carries the administrative suffix alongside the name variants, a prefecture supplied without its suffix, in romanised form, or without macrons resolves to the same record.
The Japanese address vocabulary is maintained as reference data, so the words that identify each level are managed centrally rather than being interpreted differently by each consumer.
Levels below the prefecture are preserved as submitted. Where an address provider is enabled, it can enrich and complete them.
Properties of the Japanese model worth knowing before designing an integration.
| Property | What it means for an integration |
|---|---|
| Validation stops at the prefecture | Rural districts, municipalities, wards and neighbourhoods are accepted and preserved, and are not yet checked against a CDQ list. They are never discarded. Where an address provider is enabled, the result for these levels comes from that provider. |
| The neighbourhood is not available as a separate value | 町域 is delivered inside the thoroughfare, in front of the block number, rather than in its own field. It is never lost, but a consumer that needs the neighbourhood on its own has to read it off the front of the street value. This is the deliberate consequence of giving the ward its own locality level, explained under the rejected alternatives above. |
| Read the ward from its locality level | A designated city's ward is locality level 2 and reaches the district output field. The level is what the output mapping reads and is the reliable way to find the ward. Do not discriminate on the locality type instead, and do not infer the ward from an input column's name. |
| Block-number notation is not normalised automatically | Kanji, dash and mixed notations are all accepted, and are not currently rewritten into a single canonical form. Expect to see all three in stored data. |
| Building, floor and room may arrive combined | Where the source supplies them as one string, they may be stored as a single untyped premise rather than as three typed entries. Do not assume the BUILDING, LEVEL and ROOM types are always populated separately. |
| Kyoto and rural patterns fall outside the model | Kyoto street-name references and 大字 / 字 parcel addresses do not decompose into the chome-banchi-go structure. They are preserved rather than parsed. |
Below the prefecture, the enriched result comes from the address provider rather than from CDQ reference data, and providers answer inconsistently for Japanese input. Where your own Japanese master data is maintained to a known standard, compare the two before adopting provider values, and keep the original available. This is the same advice that applies to any country where the provider, not CDQ, is the authority for a level.
Because Japan has several levels that a western model has no slot for, and because romanised addresses collapse them, a field named city may hold a prefecture, a city, a ward or a combination of all three. Read the administrative marker (都, 府, 県, 市, 区, 町, 村, 郡, or Prefecture, City, Ward in romanised data) and treat that as the authority, not the field the value arrived in.
A Japanese address is area-based rather than street-based: a nest of administrative areas from the prefecture down to a neighbourhood, then a three-part block number that locates the building. There is no street name, and the block number is a single unit that must not be split.
CDQ places the prefecture in administrative area level 1 and a rural district in level 2. The municipality goes in locality level 1, and a Tokyo special ward counts as a municipality because that is what it is. A designated city's ward goes in locality level 2. The neighbourhood and the block number travel together in the thoroughfare value, so the street line reads 丸の内1-9-2 rather than a bare number, and the house number is deliberately left empty. Building, floor and room are typed premises.
All 47 prefectures are maintained as CDQ reference data with their administrative suffixes, ISO codes and name variants, so a prefecture written with or without its suffix, in kanji or romanised, with or without macrons, resolves to the same record. The Japanese address vocabulary is maintained as reference data: the level markers, the block-number components and the premises words. Levels below the prefecture are preserved as submitted and are not yet validated against a CDQ list.
- Business partner metadata for Japan
- Japan Post postal code service
- Korea Address Standard
- Thailand Address Standard
We are constantly working on providing an outstanding user experience with our products. Please share your opinion about this tutorial!