Why Rental Data Was Always a Guess — And What Changes When It Isn’t
For as long as there’s been a rental market, there’s been a version of this question: is the number on the listing actually the number? Ask anyone who’s built a pricing model, a forecasting tool, or a CPI estimate on top of rental data, and you’ll get the same answer — probably not, and there was never a good way to check.
That question opened a recent episode of The Innovation Blueprint Podcast, ORIL’s series on emerging PropTech and the people building it. Roman Havrylyuk, ORIL’s CEO, sat down with Jonas Bordo, co-founder and CEO of Dwellsy, for a conversation that ended up covering less about any one product and more about a set of problems most of the industry is quietly working through right now: what “accurate” data even means, who’s responsible for making it AI-ready, and why building a working prototype has gotten so much easier while building something production-safe hasn’t.
Three threads from that conversation are worth pulling out, whether you’re a PropTech founder, a data or BI expert, or anyone deciding how to get from an idea to something clients can actually rely on.
The trust problem in rental data
Bordo’s company, Dwellsy, sits at an interesting intersection: a free rental marketplace on one side, and on the other, a data business selling verified rent figures to economic researchers, multifamily operators, banks, insurers, hedge funds, and other proptechs — twelve client categories in total, by his count. What made that data business viable at all is a problem the industry has wrestled with for a long time: getting an accurate asking rent, at scale, in something close to real time.
Scraping doesn’t hold up well here, because the rental market is too fragmented and moves too fast — get a site right today and it’s changed by tomorrow, across millions of listings that would need to be re-scraped constantly. Surveys run into the opposite problem: asking a landlord what they charge doesn’t get you a number you can verify, and you can’t run that survey every five minutes anyway.
Dwellsy’s way around both was to go straight to the source: pulling from the marketing feeds that property management systems use to populate listings — the same feed that powers their own marketplace.
What stood out in the conversation wasn’t the data itself so much as where it ends up getting used. Bordo expected demand for revenue management and acquisition underwriting. He didn’t expect hedge funds trading retail stocks off rent trends, or researchers using the data to help fill a long-standing gap in how the shelter component of CPI gets estimated — shelter makes up roughly 30% of that index, and rent is a major piece of it, with no reliable real-time read on the rent side until now.
AI didn’t change the data. It changed who’s responsible for it being clean
Ask Bordo how AI has changed Dwellsy’s business and the answer isn’t about a new feature — it’s about a shift in accountability. Real estate has always run on human judgment applied to physical assets, and that’s not going away. But feeding bad data into an AI model doesn’t just produce a bad number — it produces a confidently wrong one, at speed, which is worse.
That’s a shift from how BI has traditionally worked. Cleaning and normalizing data used to be something the client’s analysts did downstream, after the fact. Roman put it directly in the conversation: once AI is in the loop, “it is expected that companies provide AI-ready data, AI-ready solutions” — because any inconsistency doesn’t just confuse a spreadsheet, it produces contradictory outputs from a model that sounds certain either way. Bordo agreed, and framed it as a responsibility question: the party closest to the data — with the data science and technical capability to actually understand it — is the one who should own getting it right. Asking a client juggling ten or twenty separate data sources to normalize each one individually isn’t reasonable. That’s infrastructure work, not analyst work.
Dwellsy pulls from more than 30 different property management systems, each with its own format and quirks. Getting that down to something usable isn’t a one-time cleanup — it’s ongoing engineering. It’s the same category of problem ORIL works through with clients building on fragmented listing, brokerage, or property management data: normalization isn’t a preprocessing step you bolt on later, it’s the foundation the rest of the product sits on.
The join-key problem nobody had solved
Combining Dwellsy’s rent data with other datasets — which clients do constantly — runs into a more specific problem: there’s no reliable key to join on. Address sounds like it should work, but in multifamily especially, it breaks down fast. No existing real estate data standard covers this at the unit level.
Dwellsy’s answer is the URU — Unique Rentable Unit — a code intended to identify every rentable home or apartment in the country down to the individual room or bed. It’s launching now, and the goal is for it to function as the join key the industry has been missing, letting Dwellsy data combine cleanly with other sources instead of relying on an imperfect address match.
It’s a narrow, unglamorous problem, and that’s exactly why it matters — most of the friction in real estate data work isn’t the flashy modeling, it’s getting disparate systems to describe the same thing the same way.
Vibe coding gets you a demo. It doesn’t get you a product.
The most useful part of the conversation, for anyone weighing how to build right now, was Bordo’s read on the last few months specifically — not the last two years. Something shifted in the first quarter of this year. Building a working prototype no longer requires a technical co-founder. Dwellsy has fielded three inbound conversations in a single week from people building proptech companies with no technical partner at all, just AI tools doing the coding.
Bordo’s take on this is worth sitting with, because it’s more measured than the usual “AI writes your app now” framing: he’s genuinely excited about the door this opens for non-technical founders to prove out an idea fast. But he draws a hard line between that and what comes next. “There’s a big difference between a vibe-coded application and something that is production ready and stable and safe for clients to use with their data,” he said. “That last mile is going to be really important for people to keep in mind when they start building.”
That gap — between a prototype that proves the concept and a system a real client will trust with their data — is where most early-stage proptech ideas stall out. It’s also, not coincidentally, the exact point where a development partner earns its keep: hardening a proof of concept into something with proper data architecture, security, and the ability to scale past the first pilot client. Lowering the barrier to a first demo is genuinely good news. It doesn’t lower the bar for what “production ready” means once real client data is on the line.
Where this leaves the market
Bordo’s broader read on the industry is cautiously optimistic: more building, more supply, and a market still early in its consolidation curve. The largest property manager in the country recently crossed one million units under management — a real milestone — and it still represents roughly 2% of the 46 million rental units in the US. Scale is becoming achievable in a way it wasn’t before, largely because the tooling now exists to support it, but the industry remains deeply fragmented.
He also expects fewer one-size-fits-all platforms going forward. As AI lowers the cost of building software, he expects more tools built deliberately for a narrower audience — not a monolithic platform trying to serve everyone adequately, but something built for exactly who’s using it.
That’s a trend worth watching for anyone in proptech, brokerage, or real estate data thinking about their own tooling: the calculus on “buy the big platform” versus “build something narrower and better fitted” is shifting, and it’s shifting because building has gotten cheaper — up to the point where it needs to hold up in production.
This is the kind of conversation The Innovation Blueprint Podcast keeps coming back to: the underlying shifts in data, AI, and infrastructure that everyone building in PropTech is navigating at the same time.
Join the Conversation
Want to join the podcast? Drop us a line and share your insights on how innovation is shaping the future of real estate and PropTech!
