What is a neighborhood? 486,187 answers in our neighborhood boundary data

Ask two people where a neighborhood ends, and you often get two different answers. Ask an appraiser, an MLS, a developer, and the person who has lived on the street for thirty years, and you get four.

That disagreement is not a flaw in the question. It is the question. It is also why, as of August 26, 2026, Local Logic supports 486,187 named geographies across the United States and Canada. Neighborhood boundary data at that resolution is not an attempt to draw one correct map. It is an attempt to carry enough definitions that whoever is asking can find the one matching their geographic reality.

There is no single correct neighborhood boundary

Neighborhood boundaries are contested nearly everywhere, and the disagreement is measurable rather than theoretical. When the Bostonography project asked residents to draw their own neighborhood outlines, it collected around 300 submissions covering 21 Boston neighborhoods. Some came back clean: the North End was among the best-defined areas in the city. In Hyde Park, no part of the map reached even 75% agreement among the people who live there.

Two years ago we published a piece on who decides what a neighborhood is. The short version: appraisers define one thing, developers define another, MLSs largely decline to enforce a definition because there is no legal one to enforce, and residents hold the version that actually governs how people talk. All of those definitions are real, for different purposes.

That leaves a data company with a choice. Pick one definition and defend it, or carry as many as customers actually need. We picked the second, which means the work is unfinished by design.

What our neighborhood boundary data looks like on the ground

The August release added named residential subdivisions as neighborhood-level geographies across every US state, following a first batch in Florida, California, and South Carolina in July. US neighborhood coverage moved to 364,462.

Metro Atlanta went from 331 neighborhoods to 15,183, and Greater Boston from 363 to 3,142. The two maps below are drawn at identical frame and scale, so the change is coverage rather than zoom.

Franklin, Tennessee is a good place to see what that means in practice. Query Franklin today and Local Logic returns 161 active neighborhoods: Fieldstone Farms, McKay’s Mill, Ladd Park, Berry Farms, Cool Springs East, Stream Valley, Tywater Crossing, and 154 more. These are the names posted at the entrance to each development. They are what people say when someone asks where they live, and what agents type into a search box.

Coverage here means data, not just an outline. Hardison Hills, one of those Franklin neighborhoods, returns a full profile: mainly single detached homes alongside townhouses and small apartment buildings, most built after 2000, a grocery store within a short walk, primary schools nearby, no high school inside the boundary, several bus lines and stops, limited bike lanes. That is measured detail about one named area of roughly a square kilometre, at the depth available for a downtown Toronto neighborhood.

One address, five named geographies

A single coordinate sits inside several neighborhoods at once, and which one is correct depends on who is asking. Local Logic returns all of them, ordered from smallest to largest, and lets the caller choose. One coordinate in Franklin returns five nested geographies:

An MLS building a member search experience may want level 10, because subdivision is the field its members already file listings against. A portal building community pages for organic search may want level 10 and level 32, because those are the terms people type. A mortgage team may want the county. None of those is wrong, and none of them has to accept someone else’s answer.

This is also why a name is not an identifier. Search Local Logic for Westhaven, and you get 38 different neighborhoods across 17 states, from Franklin, Tennessee to Bethesda, Maryland to Pearland, Texas. Any platform matching a listing to a neighborhood on the name string alone will get it wrong somewhere. Geography IDs exist for that reason, and every new geography has one.

What this changes, depending on what you’re building

For portals, the useful number is not 486,187. It is how many pages you can now build that somebody is actually searching for. A metro that supported a few hundred neighborhood URLs now supports thousands, each with its own neighborhood profile and scores rather than thin duplicated text.

For brokerages, agents compete on knowing the specific area, and that is usually the subdivision rather than a city quadrant. Profiles now exist at the level of the name a client says out loud.

For MLSs, subdivision is already a field in most systems. Coverage at that level means member-facing tools can align with the geography members already use, instead of translating between two vocabularies.

For platform and data teams, this arrived without a migration. Same endpoints, same ID format, more geographies behind them. If you take bulk data delivery, the new geographies appear in your next extract.

How we keep adding definitions

The list grows for two reasons that depend on each other. Customers tell us what is missing, and the pipeline can absorb what they ask for without a project plan wrapped around every addition.

In this release, tripling the geography count meant tripling the volume of market statistics produced, and the unit cost of producing them went down rather than up. A pipeline that can only be extended by spending more stops being extended.

A few weeks ago, we published the amenity score refresh that travelled from a raw POI snapshot to live scores in 67 hours. Same instinct, different axis: that release was about how fast the data moves, this one about how many places it moves for. Both come out of one decision, which is to put the engineering into ingestion and processing rather than presentation.

We also corrected and removed geographies in this release based on specific customer requests, and that continues. A list of 486,187 named places contains mistakes. Hearing about them is how the next version gets better.