Key Takeaways
What agents actually open, why an MLS feed is a contract rather than an API, and where a real estate app budget really goes.
- NAR’s 2026 survey of 1,200 agents puts MLS use at 96% and marketing automation at 16%. Most feature budgets are spent nearer the bottom of that list.
- An MLS feed is not an API you integrate once. It is a certified data contract, and two certified MLSs can still hand you different fields.
- Photo and video delivery is the real cost center in a listing app, and it scales with inventory rather than with users.
- Any feature that surfaces neighborhood demographics is a Fair Housing surface. HUD’s 2019 charge against Facebook described exactly this mechanism.
- Only 17% of agents named the local MLS as their highest quality lead source, so a listing app is a retention tool, not a lead engine.
Start from what agents actually open
Before any feature list, read the usage data published by the people whose job you are trying to improve.
Quick Verdict: If you are scoping a real estate app, spend the first version on the MLS data contract and the photo pipeline, and leave virtual tours, lead scoring and marketing automation out entirely. Those are the features NAR’s own numbers show agents open least. A disciplined mobile app build that nails data freshness and image loading beats a feature-rich app with a stale feed. Scope it as a mobile MVP and add the showcase layer only once the data layer is boring.
NAR’s 2026 REALTORS Technology Report, covered in a NAR news article dated 22 September 2026 and based on 1,200 agent responses, ranks self-reported tool use like this.
| Tool | Agents using it |
| MLS | 96% |
| E-signature | 79% |
| Showing schedule tools | 68% |
| CMA and pricing tools | 59% |
| Drone photography and video | 48% |
| CRM | 46% |
| AI-generated content | 41% |
| Transaction management | 36% |
| Virtual tours | 29% |
| Personal website using IDX | 26% |
| Lead generation | 21% |
| Marketing automation | 16% |
Read the top and the bottom together. The MLS sits at 96%, which means the data is the product. Virtual tours sit at 29% and marketing automation at 16%, which are the two things most pitch decks lead with.
The same report says 81% of agents prioritize tools that save time and 71% look for tools that improve the client experience. Neither of those is a feature. Both are properties of a fast, current, reliable data layer.
Do this: Take your feature list and write NAR’s usage percentage next to each item. Cut anything under 30% from version one.
An MLS feed is a contract, not an API
This is where real estate app timelines break, and it breaks for a reason most teams do not see coming.
The Real Estate Standards Organization publishes the standard that makes listing data portable. RESO describes its Data Dictionary as the real estate industry’s universal language for data, and says it ensures fields and pick lists in real estate listing data are consistent across tools. Data Dictionary 2.0 is current, with 2.1 in draft.
Here is the sentence that changes your estimate. RESO states that adoption of the full Data Dictionary is not expected for most organizations and is not required for certification.
So two MLSs can both be certified and still give you different fields. Certification tells you the feed speaks the language. It does not tell you the feed contains the vocabulary your app needs.
RESO also certifies MLSs on the RESO Web API, and says MLSs certify Data Dictionary standards on the RESO Web API. That is the modern path, and it is far better than what came before. It still leaves you negotiating per market.
RESO publishes several other standards that matter once you pass one market. There is a Common Format used to certify standalone Data Dictionary payloads, a multi-language metadata standard that lets users choose a display language, and a Unique Identifiers standard. That last one is the quiet one. If you operate across overlapping MLS territories, the same physical property can arrive twice through two feeds, and deduplication is your problem rather than theirs.
What that means for a multi-market app
Every new MLS is a new approval, a new set of rules, and a field-mapping job. Budget it as onboarding work per market, not as one integration. The work resembles any third party data integration more than it resembles pulling a public API, and the slow part is paperwork rather than code.
NAR’s 2025 REALTORS Technology Survey, fielded in July 2025 with 1,241 usable responses from 49,233 invitations and a margin of error of plus or minus 2.78% at 95% confidence, reported median satisfaction with MLS core technology at 3 on a 1 to 5 scale, which the report labels content and neutral. Agents are not delighted by the systems they depend on. That is your opening.
Do this: Before writing code, request the RESO certification status and the actual field list for every MLS in your launch markets. Compare the lists side by side.
Photos are the cost center nobody budgets for
A listing app is an image delivery product wearing a search interface.
Photo volume scales with inventory, not with users. A single market with 8,000 active listings at 30 images each is 240,000 source images, and every one needs several derivatives: a list thumbnail, a card image, a full-screen view, and something for a share preview. Add drone media, used by 48% of agents per the 2026 report, and the file sizes climb again.
That changes three things in your architecture.
First, storage and egress become a recurring line item that grows with the market count. Second, the gallery becomes your hardest performance problem, because a scroll through 30 full-resolution photos on a weak connection is the moment users leave. Third, you need an ingestion pipeline that handles the feed replacing images without warning, because listings get rephotographed.
Teams that treat images as an afterthought ship version one, watch it buckle, and then face the rebuild they could have avoided. The fix is unglamorous: derivative generation at ingest, aggressive caching, a CDN, and a gallery that loads one image ahead rather than all of them.
Your framework choice matters here more than anywhere else, because image-heavy scrolling is where cross-platform tooling shows its seams. The React Native and FlutterFlow trade-off is decided by the gallery, not by the forms.
Do this: Multiply your launch market’s active listing count by 30, then by the number of derivatives you need. Price storage and egress on that number before you quote the project.
Features that surface neighborhood data are a compliance surface
This is the section competitors leave out, and it carries the most risk per line of code.
The Fair Housing Act applies to the advertising of housing. In May 2024 HUD issued guidance on how it applies when algorithms are involved, stating that the Fair Housing Act applies to tenant screening and the advertising of housing, including when artificial intelligence and algorithms are used to perform these functions. HUD also said that housing providers, tenant screening companies, advertisers, and online platforms should be aware of it.
The clearest illustration is older. In HUD News Release 19-035, dated 28 March 2019, HUD charged Facebook with housing discrimination. HUD alleged the platform enabled advertisers to exclude people whom Facebook classified as parents, non-American-born, and non-Christian, and to exclude people based on their neighborhood by drawing a red line around those neighborhoods.
Read that last phrase again, then look at your feature list.
A map that lets a user exclude areas is a drawing tool. A map that recommends areas based on who lives there is something else. HUD went further in that charge, alleging that Facebook uses machine learning and other prediction techniques to classify and group users, and that this may recreate groupings defined by their protected class even when advertisers seek a broad audience.
Where the line actually falls
I am not your attorney and this is not legal advice. What I can tell you is which features generate the questions. Neighborhood demographic overlays, school rating widgets presented as quality judgments, “find your people” matching, and any recommendation engine trained on who clicked what in which zip code. All four are buildable. All four need a fair housing review before they ship, not after.
Do this: List every feature that touches geography plus a human attribute. Send that list to a fair housing attorney before the sprint starts.
The one AI feature agents already use is the listing description
Every proptech pitch in 2026 has an AI section. NAR’s data says only one of those features has landed.
The 2026 report puts AI-generated content use at 41% of agents, and among the agents who use AI at all, writing listing descriptions is the top use at 75%. Lead prioritization is far behind, at 11% of AI users. Among emerging technologies, 48% say they are watching AI-powered lead generation and follow-up, which is a different thing from using it.
So the pattern repeats. The AI feature that gets opened operates on the listing you already have. The AI features that get pitched try to predict which buyer matters.
If you build description generation, it is a contained job: the structured fields are already in the feed, so a text generation integration has clean inputs and a human approves the output before it publishes. That last clause is not optional.
An AI-written listing description is housing advertising. HUD’s May 2024 guidance says the Fair Housing Act applies to the advertising of housing including when artificial intelligence and algorithms are used to perform these functions, and the same release warns that violations may occur when certain ad targeting and delivery functions unlawfully deny consumers. A model that cheerfully writes “perfect for young families” has just introduced familial status into your advertising copy.
My working model: ship generation with mandatory human review and a blocked-phrase list, or do not ship it. The review step is what separates a time-saving tool from an automated compliance problem.
Do this: If description generation is in scope, write the blocked-phrase list before the prompt, and make publishing require a human tap.
Scope the first version around the three things that get opened
Version one of a real estate app has three jobs and they are all boring.
Search that reflects the feed accurately. A gallery that loads fast on a weak connection. Data that is current enough that an agent is never embarrassed in front of a client.
That is it. Everything else is version two.
| Build in v1 | Defer to v2 | Reason |
| MLS feed with verified field mapping | Additional markets | Each market is an approval, not a config change |
| Image ingest with derivatives and CDN | Virtual tours | Tours sit at 29% use |
| Saved searches and alerts | Lead scoring | Lead generation tools sit at 21% use |
| Showing request handoff | Marketing automation | Lowest use in the 2026 report at 16% |
| Agent and listing attribution | In-app chat | Attribution is usually an MLS requirement |
The 2025 survey gives the commercial reason to sequence it this way. Asked which technology produced their highest quality leads, agents named social media at 39% and CRM at 23%, with the local MLS at 17%.
So a listing app is not a lead engine. It is a retention and credibility tool for agents who already have the relationship, which is a smaller promise and a far more defensible one. We publish what our builds actually delivered rather than projecting lead volume, because projecting lead volume for a listing app is not honest.
Do this: Write your version one scope as three sentences about data, images and freshness. If a fourth sentence appears, move it to version two.
What an app cannot fix, and what the MLS will not allow
Some of this is outside your control, and pretending otherwise burns a quarter.
You do not control the feed’s quality. If an MLS publishes inconsistent pick lists, your app displays inconsistent pick lists. You can normalize and you can flag, but you cannot invent data that the source does not carry.
You do not control approval timelines. Each MLS decides whether your use case qualifies, and that decision is not a technical one.
You also cannot assume the field list is stable. RESO publishes versions, 2.1 is in draft, and an MLS can change what it exposes between your estimate and your launch.
I could not retrieve NAR’s IDX rules page while writing this, so this piece deliberately makes no claim about specific IDX display requirements, attribution text, or excluded-listing rules. Get those from your MLS’s own participant agreement rather than from any article, including this one.
Where we fit: if you have one MLS approved and need a first version that handles the feed and the photo pipeline properly, that is a contained three week build and we will tell you honestly if your market count makes it longer. We are the wrong call if you have no MLS approval yet, because then your blocker is paperwork and no developer shortens it. Send us your field list and market count and we will tell you which of the two you are. You can see how we scope and sequence builds before you talk to anyone.
Do this: Confirm MLS approval in writing before you commission a single screen.
Frequently Asked Questions
How much does real estate app development cost?
Cost is driven by market count and image volume, not screen count. Each additional MLS adds an approval, a field-mapping job and ongoing feed monitoring, and image storage plus egress grows with active listings rather than users. Price one market and one feed first, then treat every further market as a separate line item.
Can I just use a public listings API instead of the MLS?
Not for current, complete market data. MLS participation is governed by agreements with each MLS, and RESO certification covers the data standard rather than your right to the data. Syndicated or scraped sources tend to lag, carry gaps, and create attribution problems. Start by confirming which MLSs will approve your use case.
Is RESO certification enough to guarantee the fields I need?
No. RESO states that adoption of the full Data Dictionary is not expected for most organizations and is not required for certification. Certification confirms the feed speaks the standard, not that it contains every field your app uses. Request the actual field list for each MLS and compare them before estimating.
Are school ratings and neighborhood data safe to show in a real estate app?
Treat them as a Fair Housing question, not a product question. HUD has stated the Fair Housing Act applies to housing advertising including when algorithms perform the function, and its 2019 charge against Facebook described excluding people by drawing a red line around neighborhoods. Have an attorney review any geography plus attribute feature before launch.
Should the first version be agent-facing or consumer-facing?
Agent-facing is the safer first build. NAR’s 2026 report puts MLS use at 96%, so agents already live in listing data daily and will tell you immediately when your feed is wrong. Consumer apps compete with portals that have a decade of brand recognition and far larger media budgets.
Resources
- The Tech Real Estate Agents Rely on Most in 2026, National Association of REALTORS, 22 September 2026. The 2026 REALTORS Technology Report usage figures from 1,200 agents, including MLS at 96%, virtual tours at 29% and marketing automation at 16%.
- 2025 REALTORS Technology Survey, National Association of REALTORS Research Group, fielded July 2025. Highest quality lead sources at social media 39%, CRM 23% and local MLS 17%, plus the sample of 1,241 usable responses and the margin of error.
- RESO Data Dictionary, Real Estate Standards Organization. The current Data Dictionary 2.0, the 2.1 draft, the RESO Web API certification path, and the statement that full adoption is not required for certification.
- HUD Issues Fair Housing Act Guidance on Applications of Artificial Intelligence, U.S. Department of Housing and Urban Development, News Release 24-098, 2 May 2024. That the Fair Housing Act applies to housing advertising including when artificial intelligence and algorithms perform these functions.
- HUD Charges Facebook With Housing Discrimination, U.S. Department of Housing and Urban Development, News Release 19-035, 28 March 2019. The alleged exclusion mechanisms, including excluding people based on their neighborhood by drawing a red line around those neighborhoods.
M. Sabir Junaid
Founder • CEO
M. Sabir Junaid is founder and CEO of Velcod, the senior product studio. He leads product and delivery across MVPs, SaaS platforms, and mobile apps — system architecture, no-code and low-code engineering, and AI integration. An official Bubble partner, Sabir has shipped products across fintech, marketplaces, edtech, and creator platforms. He works with founders raising pre-seed to Series A who need a product built to launch and scale in weeks, not a six-month agency build.
Share
- Previous
- Next
Ready to build?
Your idea, live in 3 weeks.
Book a free strategy call and get a fixed-price roadmap for your MVP, SaaS, or app.