Meet API v2.0 with UK support 🇬🇧 Try it out

Guides
for everything Geocodio

Guides
Accuracy Types & Scores

Accuracy Types & Scores

For every result, Geocodio will give you an accuracy type and an accuracy score. We recommend using a combination of the accuracy score and accuracy type to evaluate and filter the returned results.

How accurate is Geocodio, really?

Given a real-world sample list evenly distributed across the US — a list of Starbucks locations downloaded from AllThePlaces —you can expect 80% of results to be rooftop, 15% to be range interpolated, nearest_rooftop or point, and 5% to be other accuracy types.

In our latest test, of rooftop results, 75% came back building_centroid, 1% parcel_centroid, and 6% unit-level. 88% came back with an accuracy score of 1, 94% of results came back with >0.9, and 99% >0.7.

Accuracy types

Accuracy types for forward geocoding include:

  • rooftop: on the exact parcel

  • point: generally, in front of the parcel on the street

  • range_interpolation: generally, in front of the parcel on the street

  • nearest_rooftop_match: the nearest rooftop point if the exact point is unavailable

  • intersection: An intersection between two streets

  • street_center: A central point on the street

  • place: zip code or city centroid

  • county: county centroid

  • state: state centroid

For reverse geocoding, the following accuracy types may be found as well:

  • nearest_street: Nearest match for a specific street with estimated street number

  • nearest_place: Closest city/town/place

For UK addresses, the same accuracy types apply. Great Britain data is sourced directly from Royal Mail and the Ordnance Survey.

In Northern Ireland, results are currently capped at postcode-level accuracy rather than rooftop.

Match types

All geocoding results include a match type field that indicates the type of coordinate match. This field provides additional granularity beyond accuracy type and can help you understand the placement of the returned coordinates.

In the US, approximately 57% of rooftop results will come back with building_centroid-level match, 3% parcel_centroid, and 40% null.

building_centroid Coordinates derived from a building footprint centroid. Pairs with rooftop and nearest_rooftop_match accuracy types
parcel_centroid Coordinates derived from a property parcel centroid. Pairs with rooftop and nearest_rooftop_match accuracy types
unit Unit-level coordinates from secondary address line geocoding. The result has been matched to a specific apartment, suite, or unit within a building
null No specific match type information is available. This can apply to any accuracy type, including rooftop results where the coordinate source type is not known, as well as range_interpolation, street_center, place, and other result types

Unlike accuracy_type, the match_type value can be null when no specific match type information is available.

Accuracy scores

Accuracy scores are a reflection of the amount of differences between the input and the output. We generally recommend using results with an accuracy score above 0.8. Results below that threshold can indicate potential issues, such as formatting issues or incomplete addresses.

  • 1: the exact input was returned

  • 0.8: Very close to the input with minor changes made

  • <0.6: More significant changes made; use these results with caution

US Example: 47 N State St, Lake Oswego, OR 97034

  • The full address

    • Accuracy type: rooftop

    • Accuracy score: 1

  • Incomplete address (remove a digit from the ZIP code)

    • Accuracy type: still rooftop

    • Accuracy score: 0.9, reflecting that we had to change the ZIP input to arrive at the result

  • Remove the street address and geocode just the city, state

    • Accuracy type: place (i.e., a centroid for the city)

    • Accuracy score: 1, as the city/state names were spelled correctly

UK Example: 10 Downing St, London SW1A 2AA, UK

  • The full address

    • Accuracy type: rooftop

    • Accuracy score: 1

  • Incomplete address (remove a digit from the ZIP code)

    • Accuracy type: still rooftop, reflecting that we had to change the ZIP input to arrive at the result

    • Accuracy score 0.95

  • Remove the street address and geocode just the city and country

    • Accuracy type: place

    • Accuracy score: 1, as the city/country names were spelled correctly

See more about our data sources here.

Detailed accuracy explanations

Enterprise customers can optionally add the verbose parameter when using the API, which gives a more detailed breakdown of the reasoning behind the accuracy score and type. See more in our Enterprise API documentation.

Have an accuracy issue?

If you have an accuracy issue with a particular address, please submit an address issue report. This will put the address directly into our engineering queue for root-cause investigation, and you'll receive updates on our investigation.

You can submit an address report via your dashboard (requires login):