Testing location data with edge cases, not just happy paths
An address in a well mapped city center tells you almost nothing about how your system handles a rural route, a disputed border, or a query near the poles. Test the hard cases deliberately.
Latitude and longitude are not absolute measurements floating free of any model. They are coordinates on a specific mathematical approximation of the Earth's shape, called a datum, and the numbers you get depend on which datum was used to compute them. Two different datums can produce noticeably different coordinates for the exact same physical point on the ground, which is why the choice of reference system is not a minor technical footnote.
WGS84, the World Geodetic System defined in 1984, is the reference system almost universally used today for GPS and consumer web mapping. It became the default largely because GPS itself was built around it from the start, and once GPS became the dominant way both devices and services determine location, WGS84 effectively became the shared language that everything else needed to speak to interoperate. A coordinate pair with no stated reference system is, by strong convention today, assumed to be WGS84 unless something specifically says otherwise.
This matters practically any time you combine coordinates from different sources. Older surveying data, certain national mapping systems, and some legacy datasets use different datums, and plotting their coordinates directly on a WGS84 basemap without converting first will place points in the wrong spot, sometimes by a meaningful distance depending on the region and the specific datums involved. The error is not random noise, it is a systematic offset from mixing coordinate systems that were never meant to be compared directly.
For the overwhelming majority of modern applications, this is a non-issue precisely because WGS84 has become the default everywhere that matters: GPS devices, phones, mapping APIs, and geocoding services including this one all use it as the assumed standard. You generally only need to think about datum conversion when working with older survey data, certain specialized national grid systems, or data explicitly labeled with a different reference system.
The practical habit worth building is simple: when you receive coordinates from an unfamiliar source, especially an older dataset or a specialized national system, check whether it states its reference system before assuming it matches everything else you are working with. When it is silent on the matter and the data is modern, WGS84 is far and away the safest assumption. Every coordinate My Geocode returns, across forward and reverse geocoding, postal lookups, and elevation, uses WGS84, consistent with GPS and virtually every web mapping platform you are likely to plot it on.