We have become very good at identifying properties. The harder problem is teaching machines which properties actually belong together.
Real estate does not have a data shortage.
We have listing feeds, tax records, parcel data, transaction histories, automated valuations, maps, imagery, property characteristics and increasingly sophisticated AI capable of processing all of it in seconds.
But before any system can accurately evaluate a residential real estate market, there is a more basic question it has to answer:
What is the market?
For a condominium, the answer may begin with one building. For a townhome, it may be a specific community. For a single-family home, it may be a subdivision, gated development, waterfront enclave or another clearly defined competitive set.
A ZIP code can contain thousands of homes that have little practical relationship to one another. A radius can cross different condominium buildings, community boundaries, housing types, waterfront positions, HOA structures and buyer pools.
Two properties can be physically close and still compete in very different markets.
That sounds obvious to an experienced real estate professional. It is considerably harder to encode into data.
We standardized the property. The market around it is harder.
The real estate industry has made enormous progress toward data standardization.
The Real Estate Standards Organization's Data Dictionary gives the industry common definitions for property data exchanged across systems. It includes a SubdivisionName field defined as a neighborhood, community, complex or builder tract, and that field is widely adopted among organizations represented in RESO's current data. (RESO Data Dictionary)
That work is essential. A common language allows systems to exchange information far more consistently than they could when every MLS and technology provider spoke a different dialect.
But a standardized field and a normalized market entity are not quite the same thing.
A field can tell a system that a listing contains the text “Subdivision X.” It does not automatically resolve every variation of that community's name, define its boundaries, establish whether it contains multiple property types or phases, understand its relationship to a master community, or determine which properties inside and outside it genuinely compete.
RESO itself continues working on this broader geography problem. In 2026 it identified geographic information as one of the industry's highly requested areas of standardization and formed a Schools & Geographies effort that specifically includes subdivisions, cities, counties and postal codes, with goals of reducing duplication and inconsistency across markets. (RESO)
That tells us something important.
Real estate has made major progress standardizing what a property is. There is still work to do standardizing what market that property belongs to.
We think that distinction will matter enormously in the AI era.
Property identity and market identity are different problems.
Property identity asks questions such as: What is this property? Where is it? Which parcel or unit is it? What are its physical characteristics? What transaction history belongs to it?
Those are fundamental infrastructure problems, and sophisticated data companies have invested heavily in solving them.
But residential market analysis requires another layer.
We call it market identity.
Market identity asks: Which residential community does this property belong to? Which properties represent genuine alternatives for the same buyer? Which sales deserve comparison? Which nearby properties should be excluded? Does the market stop at the building, extend to several competing buildings, or divide again because certain units occupy materially different positions within the same tower?
Property identity tells us what the property is.
Market identity helps tell us what the property belongs with.
That second question is where a surprising amount of real estate judgment lives.
The nearest property is not necessarily the most relevant property.
Traditional real estate analysis often begins with geography.
Search within one mile. Look in the same ZIP code. Find the nearest recent sales.
Those techniques are useful shortcuts, particularly when better market structure is unavailable. But distance and competition are not synonyms.
Consider two condominium residences.
One comparable may be twenty feet away across the hallway, but face an entirely different direction, have a different layout and overlook a parking structure.
Another may be fifteen floors higher on the same line, with the same layout and exposure.
Which one tells us more about the subject property?
The answer cannot be derived from straight-line distance alone.
High-rise real estate makes the problem especially visible because the market has a Z-axis. Floor, stack or line, exposure, view, corner position, layout, size and condition can all affect the strength of the relationship between two residences.
But the principle extends well beyond condominiums.
A gated subdivision may compete differently from homes immediately outside its entrance. A townhome section and a single-family section may share a development name while attracting different buyers. Two adjacent waterfront communities can have materially different housing stock and competitive dynamics.
Experienced professionals make these distinctions constantly.
Software needs a way to understand them too.
This is where AI can be remarkably fast—and still wrong.
Artificial intelligence can ingest twenty comparable sales and analyze them in seconds.
It can calculate averages, summarize trends, explain pricing differences and write an impressive market narrative.
There is still one question the model cannot afford to skip:
Were those the right twenty properties?
If the competitive set is poorly defined, sophisticated AI can simply produce a sophisticated explanation of the wrong market.
It can mix communities that do not compete.
It can compare different product types.
It can treat proximity as relevance.
It can analyze an entire ZIP code when the real market is one condominium building.
It can confidently explain a dataset that should never have been assembled that way in the first place.
The industry's AI infrastructure is advancing rapidly. Cotality, for example, launched AI-ready property intelligence and a Model Context Protocol server this year, explicitly describing AI-ready data as a way for models and agents to understand the meaning behind property information. (Cotality)
We agree with the larger premise: AI needs more than raw records. It needs structured context.
Our focus is on one particular piece of that context that we believe deserves considerably more attention in residential real estate:
the competitive market around the property.
AI can read the data.
It still needs to know which market the data belongs to.
A comparable sale is not just a record. It is a relationship.
This changes how we think about comparable properties.
A comparable sale is normally presented as a row in a database: address, sale date, sale price, bedrooms, baths and square footage.
But what makes it useful is not the row.
It is the relationship between that property and the subject property.
Property A may have a strong relationship with Property B because they share the same building, line, floor band and exposure.
Property C may be nearby but only weakly related.
Property D may be farther away but represent a much closer substitute in the eyes of a buyer.
Seen this way, comparable selection is not merely a filtering problem.
It is a relationship problem.
That is the thinking behind Subdivisions.com's emphasis on defined residential communities and True Apples-to-Apples™ comparability.
The objective is not to produce the largest possible pile of comparable sales.
It is to determine which available evidence is most relevant.
Sometimes intelligence means knowing what to include.
Sometimes it means knowing what to leave out.
More data is not automatically better intelligence.
The technology industry understandably celebrates scale.
More records. More attributes. More models. More computing power.
But residential market analysis has an uncomfortable counterpoint:
Adding irrelevant data can make an analysis worse.
Twenty loosely related sales do not necessarily provide a better picture than five highly relevant ones.
A ZIP-wide median may be statistically correct and still tell a particular condo owner very little about what is happening inside their building.
A one-mile radius may contain plenty of data while combining several markets that buyers experience very differently.
This is why we believe relevance needs to come before volume.
The question is not simply, “How much property data can we analyze?”
It is:
“Which property data belongs in this analysis at all?”
That is a different optimization problem.
People don't experience homeownership as a ZIP code.
There is also a consumer reason to care about this distinction.
People frequently identify their home through the community they actually experience: their condominium building, subdivision, gated community or development.
That is where the neighbor listed.
That is where another residence sold.
That is where inventory changes.
That is where a larger unit might become available.
That is where an owner may decide to downsize without leaving the place they already love.
Administrative geography remains useful. But the homeowner's question is often far more specific than, “How is my ZIP code doing?”
It is:
What sold here?
What is competing with me here?
What actually compares with my property?
Is inventory changing in my community?
If I want more space or less space, is there an opportunity here?
That is why Subdivisions.com starts with the residential community.
Not because ZIP codes, cities or radius searches have no value, but because many residential decisions require a more precise definition of the market.
The listing is not the market.
For the last generation of real estate technology, the listing became the dominant unit of consumer experience.
Search for listings. Filter listings. Save listings. Contact someone about a listing.
That model transformed real estate discovery.
But a listing tells us what is available.
It does not, by itself, define the market around the property.
The next generation of real estate intelligence has an opportunity to go deeper: from records to relationships, from location to competitive relevance, and from property identity to market identity.
That matters to homeowners.
It matters to agents.
It matters to valuation systems.
And increasingly, it matters to AI.
Structure before intelligence.
At Subdivisions.com, we believe the sequence should be straightforward.
First, define the residential market.
Then determine the relevant property relationships.
Then organize the market evidence around those relationships.
Only then should technology calculate, compare, summarize and reason about what the data means.
The architecture looks something like this:
Property Identity → Market Identity → Relevant Relationships → Market Intelligence → Decision Support
Professional judgment still matters after all of that.
No dataset completely captures interior condition, renovation quality, presentation, seller objectives, buyer psychology, negotiation dynamics or the circumstances surrounding an individual transaction.
Technology should not pretend otherwise.
The goal is to give humans—and increasingly AI—a better starting point.
Real estate spent decades digitizing the property. The next challenge is structuring the market around it.
That is the opportunity we see.
Not another database simply because the world needs more property records.
Not another AI wrapper placed over loosely organized real estate data.
Not another way to draw a radius around an address and call everything inside it comparable.
The harder and more useful problem is establishing the relationships that give residential property data meaning.
Which community?
Which market?
Which competition?
Which comparable?
Which evidence actually belongs?
Before AI can understand a residential market, somebody has to define it.
That is where Subdivisions.com begins.
Define the market first.
Subdivisions.com — market intelligence, community by community.
Comments