When we normalize a Feature, we are scaling its base value by some other quantity. For example, converting the raw counts of nearby grocery stores to a per capita estimate, by dividing the number of nearby grocery stores by the population within the same area.

In general normalizations make it easier to compare two places to one another in an apples-to-apples fashion by controlling for variables like area and population.

Per unit area

Counts or intersecting area per unit area (in square kilometers) of the boundary. This normalization is useful when you want to understand how the density of a feature in one area compares to another area. For example, understanding the density of parks in one neighborhood and comparing to a citywide average.

This is indicated in the Feature name with the per_sqkm suffix, for example poi_is_grocery_store_count_per_sqkm.

Per capita

Feature normalized (counts, areas, etc.) per person by dividing the original value by the number of people estimated to live in a given boundary. Per capita normalization is useful for understanding markets - for example determining whether an area has a low number of grocery stores, relative to the number of people living there.

This is indicated in the Feature name with the per_capita suffix, for example poi_is_grocery_store_count_per_capita.

Percent overlap / intersecting

The proportion of a given boundary occupied by a given object. Percentages are useful when comparing how one area’s usage compares to another - for example being able to find areas that have unusually high park coverage nearby.

This is indicated in the Feature name with pct_area, for example river_pct_area_intersecting_boundary.