> ## Documentation Index
> Fetch the complete documentation index at: https://www.integrate.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# ETL: NAME_MATCH_KEYS

> Generate matching keys for organization names in Integrate.io ETL, so two sources can be joined on likely matches instead of a cross join.

## Description

Returns space-separated matching keys for an organization name. Two names that are worth comparing share at least one key, so you can join two sources on their keys and score only those candidate pairs with [FUZZY\_MATCH](https://www.integrate.io/docs/etl/fuzzy-match.md) and the `company` method, instead of comparing every row to every row with a cross join.

The keys are:

* the name's distinctive words, without legal suffixes (Inc, LLC, Ltd, GmbH), entity-type words (Capital, Fund, Holdings) or very common words (Investment, National)
* its acronyms, both stated (`(ADIA)`, `APG - All Pensions Group`) and derived from its words. A parenthetical tag that doesn't spell the name's initials, such as `(UK)`, isn't a key
* the whole normalized name, prefixed with `n:`, so identical names always share a key

Keys are lower case and sorted, so the same name always produces the same keys.

## Syntax

`NAME_MATCH_KEYS(string_expression)`

## Examples

`NAME_MATCH_KEYS('ADIA')` returns `'adia n:adia'`.

`NAME_MATCH_KEYS('Abu Dhabi Investment Authority (ADIA)')` returns `'abu adi adia dhabi dia n:abu_dhabi_investment_authority'`. It shares the key `adia` with `'ADIA'`.

`NAME_MATCH_KEYS('Temasek Holdings Pte Ltd')` returns `'hpl n:temasek_holdings_pte_ltd temasek thp thpl'`. It shares the key `temasek` with `'Temasek'`.

`NAME_MATCH_KEYS('H&M')` returns `'hm n:hm'`. Letters that spell a name are kept together, so `'H and M'` and `'HM'` share the key `hm`.

To produce one row per key, ready to join on, use it with `TOKENIZE` and `FLATTEN` in a Select component:

```sql theme={null}
FLATTEN(TOKENIZE(NAME_MATCH_KEYS(company_name), ' '))
```

## Return value datatype

String.

## Notes

* A name with no usable words, such as `'!!!'`, an empty string or a placeholder like `'N/A'`, `'Unknown'` or `'TBD'`, returns `null`, so blank names don't all share a key and multiply in a join. To keep those rows on the side whose unmatched rows you want in the output, use `COALESCE(NAME_MATCH_KEYS(name), '-')`. `-` is never a real key, so the row joins nothing and comes through a left join unmatched.
* In a mixed-case value, upper case is used to recognize acronyms and US state codes (`NJ` becomes `new jersey`), so pass the original value rather than `LOWER()`. In an all-caps value such as `ACME CO`, two-letter words are kept as words.
* Keys are meant for joining, not for display. A shared key only means the pair is worth scoring; use `FUZZY_MATCH(name1, name2, 'company')` to decide whether it's a match.

## Impact of null value

Returns `null` if the input is null, has no usable words, or is a placeholder such as `'N/A'`.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.