Geocoding

Geocoding converts a text description of a place, usually an address, into coordinates. Reverse geocoding does the opposite and returns the nearest address for a pair of coordinates. It is often the first step in a project, because most useful data arrives as a list of addresses rather than as geometry.

The difficulty is that addresses are written by people. Abbreviations, missing house numbers, old street names, spelling variants and inconsistent formats all reduce the share of records that match. A geocoder returns a confidence level along with each result, and the honest workflow is to check the low confidence matches by hand rather than mapping everything and hoping.

Two decisions are worth stating in any project that relies on geocoding: which geocoder and reference dataset you used, and what your match rate was. A dataset where a fifth of records failed to geocode is not a dataset with a fifth less data. It is a dataset with a systematic gap, and the gap is often concentrated in newer or more rural areas.