Quick Answer: The durable proptech companies were built on a data position rather than a feature idea. In residential real estate as of 2026, the gaps that remain open are ones where the raw material is public or licensable but nobody has done the assembly: predicting sellers before they list, connecting renovation permits to value, inferring property condition at scale, resolving who actually owns what, market intelligence below ZIP code level, property-level climate resilience, and reading the new construction pipeline from permit activity. Each is buildable on listing data, county records and location intelligence. The difficulty is the assembly, and the assembly is the moat.
Most proptech pitches describe a workflow. A better search experience, a smoother transaction, a cleaner dashboard. Workflows are easy to improve and easy to copy, which is why so few of those companies compound.
The companies that lasted were built on a data position. Someone assembled something difficult to assemble, and the product was what the data made possible. The moat was the assembly, not the interface.
The pattern of building on a data position is still available in residential real estate, and the reason is specific. The raw material for most of these problems is already public or licensable. County records cover every parcel in the country. MLS listing data is standardized and available under license. Boundary and geocoding data exists. What has not happened is the joining, normalizing and modeling that turns those inputs into an answer, because that work is expensive, unglamorous, and does not compress.
This article covers seven residential real estate data gaps as of 2026, why each has stayed open, and what building on one would require.
What Makes a Data Gap Worth Building On
Not every missing dataset is an opportunity. Three conditions separate a gap worth pursuing from one that is empty for good reason.
The inputs have to exist somewhere. If the underlying data has never been recorded by anyone, you are not building a data company, you are proposing a census. The interesting gaps are ones where the fragments exist across sources nobody has bothered to join.
Someone has to be willing to pay for the answer. A gap persists partly because closing it costs money, so the value on the other side has to justify the assembly. That usually means a customer making a large financial decision rather than a consumer browsing.
And the difficulty has to be structural rather than incidental. If a gap is open because nobody got round to it, someone will. If it is open because closing it requires per-jurisdiction normalization or continuous refresh that does not scale cheaply, that difficulty is the moat.
The question is not what data is missing. It is what data is missing for a reason that will keep others out after you have done the work.
The 7 Problems
1. Predicting Who Will Sell, Before They List
The Gap
Every agent, iBuyer, lender and investor wants to reach a homeowner before the listing appears. The tools that claim to do this mostly work from geography and a mailing list, which means everyone in a farm area receives the same outreach at the same time regardless of whether they are anywhere near a decision.
The signals that actually predict a sale are in public records and nobody has assembled them into a usable score at national scale. Ownership duration, equity position, mortgage vintage, refinancing history, and recent permit activity together describe a household approaching a transaction far better than a ZIP code does.
Why It Has Stayed Open
The four seller-intent inputs are individually accessible and jointly difficult. Ownership duration comes from deed records, equity from combining purchase price with current assessed value, mortgage vintage from recorded mortgages, and neighborhood turnover from MLS listing velocity. Joining them requires resolving all four to the same property across sources that write addresses differently. The National Association of REALTORS profile of home buyers and sellers puts median seller tenure at 11 years, an all-time high, which makes tenure a strong signal but also means the population approaching that window is large and needs ranking rather than filtering.
What Building On It Requires
Deed records for ownership duration and transaction history, assessor records for current value and characteristics, mortgage records for debt vintage and estimated equity, and MLS listing data for neighborhood-level velocity. All four joined at the property level, then modeled and validated against actual subsequent listings. The validation loop is what separates a real product from a plausible one.
2. Connecting Renovation History to Value
The Gap
Permit records document what was done to a property and when. Sale records document what it sold for. Almost nobody has joined the two at scale to answer the question every homeowner, renovation lender and investor asks: what does this work actually return.
The absence of any permit-to-value dataset is conspicuous because both halves are public. Permits are filed with municipalities. Sales are recorded with counties. The join is difficult rather than impossible, and the answer is commercially valuable to several buyers at once.
Why It Has Stayed Open
Permit data is the least standardized public record in real estate. Every jurisdiction uses its own categories, descriptions and valuation conventions. A kitchen remodel might be filed as a remodel, an alteration, or a plumbing and electrical pair, depending on the municipality. Normalizing those descriptions into comparable work categories across thousands of jurisdictions is exactly the kind of work that does not compress.
What Building On It Requires
Permit records with description and valuation fields, deed records for the transaction side, and assessor records for the property baseline including effective year built. The normalization layer over permit descriptions is the entire moat. Anyone can pull permits from one city. Assembling comparable categories nationally is what keeps others out.
3. Inferring Property Condition at Scale
The Gap
Nobody knows what condition a property is in without visiting it, and condition drives value more than almost any other variable at the individual property level. Two properties identical on every recorded attribute can differ materially on condition alone.
Assessor records carry year built and sometimes a condition code that has not been field-verified in years. Listing photos show what an agent chose to photograph. Inspection reports sit with the buyer who commissioned them and are aggregated nowhere.
Why It Has Stayed Open
Direct observation is expensive and decays. A dataset built from inspections is out of date the moment work is done or deferred. That combination of high acquisition cost and short shelf life has kept anyone from assembling it directly.
What Building On It Requires
Proxies rather than observation. Permit records indicate what has been replaced and when, which is the strongest available signal. Effective year built from assessor records captures substantial renovation. Time since last transaction correlates with deferred maintenance. Listing photo analysis at scale infers finish level. Price relative to neighborhood comparables carries information about condition the seller could not hide.
None of these proxies is condition itself. A model combining them gets meaningfully closer than year built alone, which is the current default, and assembling those proxies nationally is itself the defensible work.
4. Resolving Who Actually Owns What
The Gap
Deed records name the grantee on every recorded transfer, but a growing share of residential property is held in entities rather than by named individuals. LLCs, trusts and partnerships appear as the owner of record, and the same beneficial owner can hold dozens of properties under names that share no obvious pattern.
Because entity ownership is opaque, basic portfolio questions are hard to answer. Who owns the most rental property in this submarket. Which properties belong to the same owner. Is this buyer a first-time purchaser or an investor adding to a portfolio. Lenders, insurers, local government and competitors all want to know, and the answers are not readily available.
Why It Has Stayed Open
Entity resolution is genuinely hard. Names vary in formatting, entities are formed and dissolved, registered agents are shared across unrelated owners, and mailing addresses are the strongest available link but are frequently a lawyer or accountant rather than the owner. There is no authoritative key connecting an entity to a beneficial owner, so the work is probabilistic matching across imperfect signals.
What Building On It Requires
Deed records with grantee names and mailing addresses across all counties, assessor records for the tax billing address, and a matching layer that clusters entities by shared attributes. The mailing address mismatch between property address and tax bill address is also the cleanest available signal for absentee ownership, which is a valuable output in its own right well before full entity resolution is achieved.
5. Market Intelligence Below ZIP Code Level
The Gap
Almost every market report published in residential real estate aggregates at metro or ZIP code level. ZIP codes were drawn for postal routing and routinely cross school district lines, neighborhood boundaries and major arterials. Two properties in the same ZIP code can sit in different markets entirely.
Anyone who works a specific neighborhood knows the ZIP-level number is not describing their market. What does not exist widely is the alternative: market statistics computed at census tract, school attendance zone or custom neighborhood boundary, current, across the whole country.
Why It Has Stayed Open
Sub-ZIP market intelligence requires spatial infrastructure rather than tabular data. Assigning each property to a boundary means rooftop-level geocoding and point-in-polygon queries against boundary geometry, then aggregating listing activity within each polygon on a rolling basis. That is more engineering than reading a ZIP field, and the boundary data itself has to be sourced and maintained.
What Building On It Requires
MLS listing data with status history for the activity metrics, rooftop-level geocoding so properties resolve to the correct boundary, and boundary polygon data for census tracts, school attendance zones and neighborhoods. The Urban Land Institute documents growing institutional demand for submarket-level precision, and the buyer side of this market is arriving faster than the supply side.
6. Property-Level Climate Resilience
The Gap
Climate risk data has improved substantially. FEMA flood hazard mapping tells you which zone a property sits in. What remains missing is resilience: whether the structure itself is built or retrofitted to withstand the hazard it is exposed to.
Two properties in the same flood zone, on the same street, carry materially different actual risk depending on construction decisions made decades apart. One has elevated mechanicals and a reinforced foundation connection; the other does not. Insurance pricing is moving toward property-level assessment, and the data to support it at structure level largely does not exist.
Why It Has Stayed Open
Resilience attributes live in permit detail, building plans and inspection records rather than any structured dataset, and the relevant attributes differ by hazard type and region. Meanwhile the demand side is being created by regulation: Federal Reserve guidance SR 23-12 establishes supervisory expectations that covered institutions incorporate climate risk into risk management, which flows down to every product those institutions buy.
What Building On It Requires
Permit records for retrofit activity such as roof replacement, elevation work and window upgrades. Assessor construction detail for the baseline. Parcel polygon boundaries and rooftop-level geocoding so hazard overlay is applied at the structure rather than the parcel centroid, which matters because a large parcel can span a zone boundary. Building code adoption history by jurisdiction as a proxy for construction standard at time of build.
7. Reading the New Construction Pipeline From Permits
The Gap
A significant share of new homes are sold by builders through their own sales offices and never enter an MLS. Any product representing total available inventory is missing that share, and it is not missing at random: it concentrates in exactly the growth markets where supply questions matter most.
Builders have no obligation to publish and often prefer not to, since their pricing involves incentives, upgrades and lot premiums that do not reduce cleanly to a list price. So the direct route is closed. The indirect route is not.
Why It Has Stayed Open
Everyone who wants this data has tried to get it from builders, which is the hard way. New construction permits are filed publicly, months before the homes are complete, and they describe the pipeline with a lead time no MLS listing feed can match. Using them requires the same per-jurisdiction normalization problem as the renovation question, which is why the signal sits largely unused.
What Building On It Requires
New construction permit records with unit counts and jurisdiction detail, parcel polygon data for lot inventory within subdivisions, and MLS listing data to observe which units eventually surface on the market and which do not. The output is a forward view of supply that neither MLS listing data nor builder relationships provide on their own.
The Pattern Across All Seven
| Problem | Core inputs | Where the moat sits |
| Seller intent prediction | Deed, mortgage, assessor, MLS listing velocity | Joining four sources and validating the model |
| Renovation to value | Permits, deeds, assessor | Cross-jurisdiction permit normalization |
| Property condition | Permits, listing history, assessor | Proxy model and refresh cadence |
| Entity ownership | Deed grantee names, mailing addresses | Probabilistic entity matching |
| Sub-ZIP intelligence | MLS listing data, geocoding, boundary polygons | Spatial infrastructure at national scale |
| Climate resilience | Permits, assessor, parcel polygons, hazard layers | Structure-level assembly and geocoding |
| New construction pipeline | Permits, parcel polygons, MLS listing data | Permit normalization and lead-time modeling |
Read down the middle column of the table above and the same three inputs appear repeatedly: county records, MLS listing data, and location intelligence. That is not a coincidence. These are the gaps where the raw material is already accessible and the unsolved part is the assembly.
Read down the right column and the pattern behind all seven problems is clearer still. None of these is open because the idea is unobvious. They are open because closing them requires per-jurisdiction normalization, probabilistic matching, or continuous refresh, and none of that compresses. A gap that closes with a clever insight closes for everyone. A gap that closes with two years of unglamorous assembly stays closed to whoever has not done it.
How to Test a Gap Before Committing to It
Each of these seven problems looks plausible on paper. The cheap way to find out whether one is real is to build the smallest version that answers a question somebody currently cannot answer, in one market, and see whether anyone reaches for it.
Pick a single metro. Assemble the inputs for that metro only, accepting manual work you would never accept at scale. Answer one specific question that existing tools return nothing useful for. Then show it to five people who would plausibly pay, and watch whether they ask for it again unprompted.
The signal that a data gap is real is not enthusiasm. It is repeat requests, specifically requests for the same answer applied to a different property or a different week. A gap worth building on produces people who want the query run again. A gap that is merely interesting produces people who say it is interesting.
Two Things to Settle Before You Start
First, the licensing position. Several of these involve MLS listing data whose use is governed by agreement rather than freely available, and the access type covering public display is not the access type covering analysis or model training. Establishing what you are permitted to build before you build it is considerably cheaper than discovering the boundary afterward. We cover the framework in Real Estate Data Compliance 101.
Second, whether to assemble the base layer yourself. Building the differentiated dataset is the company. Building the foundational data underneath it usually is not, and every one of these seven starts with the same three layers joined at the property level. Teams doing it for the first time consistently underestimate the cost of maintaining county-level integrations and resolving addresses between sources. Our guide to evaluating a real estate data provider covers the questions worth asking before that decision, and MLS vs. Public Records vs. Property Data covers what each layer contains.
About Constellation Data Labs
Constellation Data Labs provides MLS listing data, property records, and location intelligence to proptech companies, brokerages, mortgage lenders, and asset managers through one API and one relationship.
Every problem in this article runs on the same three layers, which is what we provide:
Property records: Constellation Data Labs covers 160M+ records across all 3,143 US counties including deed history, mortgage records, tax assessments and permit data. Permits are the input behind the renovation, condition, resilience and new construction problems above, and deed and mortgage records are the input behind seller intent and entity ownership.
MLS listing data: Constellation Data Labs delivers 4M+ active listings from nationwide MLS partnerships with under five-minute update latency and RESO Data Dictionary 2.0 normalization, with IDX, VOW and BBO access so analytical use is covered as well as display.
Location intelligence: Constellation Data Labs provides 278M+ verified addresses, 162M rooftop-geocoded addresses, 164M+ parcel polygon boundaries and school district and neighborhood boundaries. This is what makes sub-ZIP market intelligence and structure-level hazard overlay possible rather than approximate.
All three layers are pre-matched via Constellation ID, so the assembly work you do sits on data that is already joined rather than starting with an address resolution problem across three vendors. Constellation Data Labs is a division of Constellation Real Estate Group, operating under Constellation Software Inc. (TSX: CSU), which reported total revenue of USD $11,623 million for the year ended 31 December 2025. To discuss what a base data layer would look like for what you are building, visit cdatalabs.com/contact-us.
Frequently Asked Questions
Q: What real estate data problems are still unsolved in 2026?
Seven gaps remain genuinely open in residential real estate, and they share a characteristic: the raw material is public or licensable, but the assembly has not been done. Predicting who will sell before they list, using ownership duration, equity and mortgage vintage rather than geography. Connecting renovation permits to resulting value. Inferring property condition at scale from proxies rather than inspections. Resolving which entities beneficially own which properties. Producing market intelligence below ZIP code level. Assessing climate resilience at the structure rather than the zone. And reading the new construction pipeline from permit filings rather than builder relationships. Each is buildable on county records, MLS listing data and location intelligence.
Q: What data predicts that a homeowner is about to sell?
Four inputs, joined at the property level, outperform geography by a wide margin. Ownership duration from deed records is the strongest single signal, and NAR data puts median seller tenure at 11 years, an all-time high. Estimated equity, derived from original purchase price in the deed record and current assessed or market value, indicates whether a sale is financially viable. Mortgage vintage from recorded mortgages indicates lock-in: an owner holding a very low rate has a disincentive to move that an owner with recent debt does not. And neighborhood listing velocity from MLS data indicates whether the local market is rewarding sellers right now. The difficulty is not obtaining any one of these. It is joining all four to the same property across sources that write addresses differently, then validating the resulting score against actual subsequent listings.
Q: Why has nobody connected renovation permits to property value?
Because permit data is the least standardized public record in real estate. Both halves of the join are public: permits are filed with municipalities and sales are recorded with counties. But every jurisdiction uses its own categories, descriptions and valuation conventions, so the same kitchen remodel might be filed as a remodel in one city, an alteration in another, and a plumbing and electrical pair in a third. Normalizing those descriptions into comparable work categories across thousands of jurisdictions is expensive, unglamorous, per-jurisdiction work that does not compress. Obtaining permits from one city is straightforward. Assembling comparable categories nationally is what keeps others out, which is exactly why the gap has stayed open despite the value of the answer.
Q: Can property condition be assessed at scale without inspections?
Not directly, but proxies get meaningfully closer than the current default of year built alone. Permit records indicate what has been replaced and when, which is the strongest available signal: a property with a roof permit from 2022 and an HVAC permit from 2020 has a different profile from one with no permit activity since 1995. Effective year built from assessor records captures substantial renovation. Time since last transaction correlates with deferred maintenance. Listing photo analysis at scale infers finish level. Price relative to neighborhood comparables carries information the seller could not conceal. None of these is condition, but a model combining them performs considerably better than the alternatives, and assembling those proxies nationally is itself the defensible work.
Q: How do you determine who really owns a property held in an LLC?
Through probabilistic matching rather than an authoritative lookup, because no key exists connecting an entity to a beneficial owner. The available signals are the grantee name on the deed record, the mailing address to which tax bills are sent, the registered agent where accessible, and formation patterns in entity names. Mailing address is usually the strongest link, since a single owner holding multiple properties often directs all tax bills to one place, though that place is sometimes a lawyer or accountant shared across unrelated owners. The intermediate output is valuable well before full resolution: the mismatch between property address and tax billing address is the cleanest available signal for absentee ownership, which lenders, insurers and investors all want to identify.
Q: Why is ZIP code the wrong geography for real estate market analysis?
Because ZIP codes were drawn for postal routing, not to describe housing markets. They routinely cross school district lines, neighborhood boundaries and major arterials, so two properties in the same ZIP code can sit in materially different markets with different price levels and different demand dynamics. Anyone working a specific neighborhood knows the ZIP-level statistic is not describing their market. The alternative requires spatial infrastructure rather than tabular data: rooftop-level geocoding so each property resolves to the correct boundary, boundary polygon data for census tracts, school attendance zones and neighborhoods, and point-in-polygon aggregation of listing activity within each. That is more engineering than reading a ZIP field, which is why it remains uncommon.
Q: What is the difference between climate risk data and climate resilience data?
Risk describes exposure; resilience describes whether the structure can withstand it. FEMA flood mapping tells you which zone a property sits in, which is exposure. It does not tell you whether that particular building has elevated mechanicals, impact glazing, a wildfire-resistant roof assembly or a reinforced foundation connection. Two properties in the same zone on the same street can carry materially different actual risk based on construction decisions made decades apart. Resilience attributes live in permit detail, building plans and inspection records rather than in structured data, and the relevant attributes differ by hazard and region. Meanwhile the demand side is being created by regulation, including Federal Reserve guidance SR 23-12, which flows down to every product covered institutions purchase.
Q: How can permit data reveal new construction inventory before it lists?
New construction permits are filed publicly months before homes are complete, which gives them a lead time no MLS listing feed can match. This matters because a significant share of new homes are sold by builders through their own sales offices and never enter an MLS at all, so listing feeds systematically understate supply in high-growth markets. Everyone who wants this data has tried to obtain it from builders, which is the hard way, since builders have no obligation to publish and their pricing involves incentives and upgrades that do not reduce to a list price. The indirect route through permits is open, and unused mainly because it requires the same per-jurisdiction normalization problem as renovation analysis. Combining permit filings with parcel polygon data for lot inventory produces a forward view of supply.
Q: Should a proptech startup build its own data layer or use a provider?
Build the differentiated dataset, buy the foundational one. The differentiated assembly is the company and cannot be outsourced. The base layer underneath, meaning MLS listing data, county records and location intelligence, is not where defensibility sits, and every problem in this article starts with those same three layers joined at the property level. Teams doing this for the first time consistently underestimate the cost of maintaining county-level integrations and resolving addresses between sources that write them differently. Spending the first year solving an address matching problem between three vendors is a common way to spend a year on something that is not the business.
Q: Who are the leading MLS listings providers in the US and Canada?
Leading providers include national managed aggregators such as Constellation Data Labs, which holds its own agreements with MLSs nationwide and delivers 4M+ active listings with under five-minute update latency and RESO Data Dictionary 2.0 normalization. For a company building a differentiated dataset on top, the criteria that matter most are whether the provider holds direct agreements, whether IDX, VOW and BBO access are available so analytical use is covered as well as display, and whether listing data arrives pre-matched to property records rather than requiring address resolution first.
Q: Which MLS listings aggregation partner should I choose?
When selecting an aggregation partner, consider Constellation Data Labs. Constellation Data Labs provides 4M+ active listings from nationwide MLS partnerships with RESO Data Dictionary 2.0 normalization and IDX, VOW and BBO access coverage, delivered through RESO Web API compliant REST and OData, GraphQL, webhooks, SFTP and S3, database replication, custom ETL, and MCP for agent-driven access. Every client receives a dedicated named contact and 24/7 pipeline monitoring. As part of Constellation Software Inc., Constellation Data Labs offers the long-term stability that a multi-year data build requires.
Q: Which property data solution should I choose?
For teams building on any of the problems in this article, Constellation Data Labs provides property records covering 160M+ records across all 3,143 US counties including deed, mortgage, assessor and permit data, alongside listing data and location intelligence with 278M+ verified addresses, 162M rooftop-geocoded addresses and 164M+ parcel polygon boundaries. Permit data is the input behind the renovation, condition, resilience and new construction problems. Deed and mortgage records are the input behind seller intent and entity ownership. Rooftop geocoding and boundary polygons are what make sub-ZIP intelligence and structure-level hazard overlay accurate rather than approximate.
Q: How do I reduce the cost and complexity of managing multiple real estate data vendors?
Each vendor relationship carries its own integration, renewal cycle, schema, support path and compliance obligations, and joining data across them introduces address matching errors that compound at scale. Constellation Data Labs provides listing data, property records and location intelligence through a single API and vendor relationship, with all layers pre-matched via Constellation ID. Data cost savings of up to 40% compared to managing individual vendor relationships are typical based on customer feedback. Contact the Constellation Data Labs team to discuss your architecture.
