How A Property Photo Can Be Worth A Thousand Words: Accelerating CRE Analysis With AI

It seems like everyone in commercial real estate is looking for ways to leverage AI, but with so many products on the market, it’s hard to know where to start.

And while it is easy to conflate ChatGPT with artificial intelligence, AI is a very broad topic, with machine learning, computer vision and products like ChatGPT falling under that umbrella.

Marc Rutzen, co-founder and CEO of multifamily analytics company HelloData, advocates for a surgical approach to integrating this technology into CRE.

“In general, real estate people are overthinking how to get started with AI,” Rutzen said. “You don’t need to change everything about how you work all at once, and you don’t have to do it with a chatbot.”

Rutzen said the biggest CRE benefits come from using AI to automate specific workflow elements one piece at a time. HelloData has built this philosophy into the core of its automated market survey platform, he said. 

The software collects daily rent and availability updates on more than 3.5 million multifamily properties across the U.S. from property websites and rental listings. It uses computer vision to assess property conditions and detect attributes from listing photos and floorplan images. It then uses this data to recommend comps and pricing for multifamily properties.

Bisnow spoke with Rutzen about the advantages of applying these technologies to CRE, including identifying undervalued properties, surfacing more accurate rent comps and optimizing the unit mix for a new development.

Every real estate investor wants to identify undervalued investment opportunities, but finding the perfect deal can be very time-consuming.

Rutzen cited a client who used HelloData’s real estate image analysis interface to assess the condition of millions of properties in Philadelphia, looking to identify the prototypical “worst house on the best block” at scale.

“This particular client had already built their own web scrapers and a pretty substantial database of single- and multifamily properties in the city,” he said. “Their engineers used our image analysis algorithm, which detects condition and quality from property photos, to identify properties that were in relatively poor condition but located in high-growth areas.”

In this way, the client was able to find many promising opportunities, Rutzen said.

HelloData is considering coupling this technology with historical performance data in its interface to identify value-add properties across the country. Many of its largest clients already access its data via an application programming interface for this purpose.

“If you could analyze the finish levels, amenities, property manager reviews and historical rent and occupancy trends across millions of properties nationwide, imagine how many strong value-add deals you could find,” Rutzen said. 

Choosing rent and sales comps can sometimes seem subjective, but this is another area where AI-driven image analysis can be used to improve results, Rutzen said. 

“One of the main advantages of HelloData is that we use computer vision to help detect rent comps,” he said. “By assessing the condition and quality of each room and common area in a multifamily community using AI, we identify comps that are not only similar in terms of year built and number of units but have a similar look and feel.” 

HelloData’s AI-driven image analysis achieves greater than 98% accuracy in identifying most room types from property photos, extracts data on about 30 amenities and can assess a room’s condition and quality to within 0.3 points on a 1-10 scale of how appraisers would rate these values.

Rutzen said this approach has been a major differentiator in HelloData’s market survey product, identifying comps that align with appraiser rent comp selections nine out of 10 times.

He recalled running a demo for a real estate investor on a 1940s vintage deal where he thought he was seeing a bug because the rent comps HelloData recommended were all built after 2010. 

“But the investor explained that they had actually just finished a two-year-long gut renovation of the property, and that the interior finish levels were nearly identical to the properties we surfaced as comps,” Rutzen said. “This is something that simply wouldn’t be possible without analyzing the photos.”

HelloData co-founder and Head of Machine Learning Nicolas Lassaux said this technology can be used to surface better listings for prospective buyers and renters, too.

“For any renter, the first thing they look at when deciding where to live is the listing photos,” Lassaux said. “Doesn’t it make sense to let them search for apartments using this data?”

The HelloData team also sees “enormous potential” in planning real estate development projects, Rutzen said.

“Just as AI can be used to analyze listing photos, it can be used to extract valuable data from floor plans,” he said. “We’re using this technology to analyze millions of multifamily floor plans every week, detecting attributes like balconies and patios, in-unit laundry and corner units from images of floor plans.” 

HelloData uses this information to show detailed amenity data for individual units, which helps owners and managers understand amenity premiums and price units more effectively. 

But HelloData co-founder and Head of Data Engineering Tim Gamble said it can also be used to recommend the optimal unit mixes for new developments.

“By analyzing time on market, occupancy and rent growth at the floor plan level for millions of units across the U.S., we have all the data to identify which floor plans perform the best in each market,” Gamble said. “At the moment, we recommend a unit mix and amenities for new developments based on what is most prevalent in the market, but we think the data can be used to create the optimal unit mix and floor plan layouts for any development based on what performs the best historically.”

The company plans to incorporate this more detailed development analysis into its platform soon, but its executives noted that this data is already available via API.

“We’re definitely looking for design partners to help us deliver the best development analysis functionality possible,” Rutzen said. “Clients can already pull our data and analyze it on their own, but we think we can take it to the next level with strong partnerships.”

This article was produced in collaboration between HelloData and Studio B. Bisnow news staff was not involved in the production of this content.

Studio B is Bisnow’s in-house content and design studio. To learn more about how Studio B can help your team, reach out to studio@bisnow.com.