The Geocoder stage looks up every address of a list in your own reference datasets, and gives each row its coordinates, the matched address and a match confidence. The data is yours; the stage brings none of its own.
Dictionary
The reference datasets the Geocoder uses for geocoding and address validation.
- Choose the Country.
- Choose Add a Dataset and pick an address layer. Or choose Copy From an Address API to start from the datasets and field mapping of an API you built in API Builder; the stage keeps its own copy.
- Under Address Fields, say which field holds each part of an address. Suggest Mapping fills in the parts it can tell from the field names, and Sample shows a value from each field; check each before you save. Map Full Address, or Street Name and the parts you have. Other street names and other community names (a highway number, a mailing city, a former name, a neighbourhood) are searched too.
- Choose the Location (the layer's geometry, or latitude and longitude fields) and, if you have one, the Record ID.
Datasets are asked in the order listed. The next dataset is asked only for an address the ones before it find nothing for, so an address that two datasets hold is answered once, from the first.
Input
Say where the address is in the rows that reach the stage:
- One Column Holds the Full Address.
- Address Lines: Address Line 1 holds the number and street together; city, state or province and postal code are in their own columns.
- Separate Parts: every part in a column of its own.
Output
Ready holds the rows where the address was found. Its one setting is Minimum Confidence: a row is Ready when its match is at the address level (the house number was found) and at least that confident. Every Ready row gains these columns, and its geometry becomes the matched point (WGS 84):
| Column | Holds |
|---|---|
matched_latitude, matched_longitude | The location of the matched address |
matched_address, matched_id, matched_dataset | The matched address, its record ID and the dataset it came from |
match_status | matched |
match_confidence | From 0 to 1 |
Need other fields of the matched record? Join the dataset on matched_id.
Needs Review holds the rows where it was not, unchanged, with the reason. Its one setting is Closest Matches Kept. Every Needs Review row gains review_reason, match_status, match_confidence and closest_matches.
match_status says what the stage found:
| Status | Means |
|---|---|
Matched (matched) | One address matched. |
Multiple Matches (multiple_matches) | More than one address matched equally well. |
No Match (no_match) | No address matched. |
| Reason | What to do |
|---|---|
| The row has no address. | Fill the address in your source, or filter such rows out before the stage. |
| No Match: no address matched. | Check the spelling, or add the address to a dictionary dataset. |
| Multiple Matches: more than one address matched equally well. | Read the closest matches and add what tells them apart: a city, a postal code, a unit. |
| Street found, house number not found. | Correct the house number. |
| The match is below the minimum confidence. | Lower Minimum Confidence, or review these rows by hand. |
| The dictionary holds no addresses for this region. | The row names a state or province your datasets do not cover. |