National CRE News In Your Inbox.
Sign up for Connect emails to stay informed with CRE stories that are 150 words or less.
AI’s Stumbling Block: Bad Data
Artificial intelligence tools, predictive analytics and other technologies are being heralded as methods to process data, gain insights and even help eliminate or greatly diminish repetitive tasks. But the issue currently facing AI’s success is the validity of its data.
A recent article in Urban Land Magazine pointed out that “commercial real estate data is not standardized or consistently high-integrity.” Furthermore, the industry “doesn’t have a lot of consistent data, and [it is] not great at making data consistent and cleaning it up,” data consultant Rick Haughey told Urban Land Magazine.
CRE data not only lacks standardization, it tends to be fragmented. The article explained that companies continue “juggling multiple, non-communicative software platforms.”
Lisa Stanly with OSCRE International noted that one organizational member collects its property data on 40 different software platforms, none of which “talk” to one another. OSCRE International is a nonprofit organization focused on developing and implementing real estate data standards. Before AI can be used effectively, investors and others must “pull the data from all these different systems and have confidence in the data quality. Even with AI, garbage data in still yields garbage data out,” Stanley explained to the magazine.
Adding to the issue is that not every CRE company has access to every data source—most real estate data tends to be privately held, which can get in the way of standardizing information. Investment professionals and brokers tend to keep as much information to themselves as possible. “That’s what makes one investment manager better than the other and the end of the day,” AJS Advisory’s Andrea Jang told Urban Land Magazine.
While the obvious solution might be to standardize the data, the article explained that this tends to be easier said than done. Faropoint launched its data efforts four years ago to prepare for training its AI models. The process involved hiring a research and development team of data engineers. Faropoint accumulated a lot of data to build a proprietary system while the team determined which parameters would impact rent (including property age and nearness to public transportation).
Even with this, Faropoint’s Adir Levitas said that creating an accurate AI model is “30% of the mission.” Additional time and resources are needed to consistently update and “integrate AI into the business processes that are going to be useful for people,” Levitas explained to Urban Land Magazine.
The upshot of the article is that better data standardization can help improve AI usage by more CRE companies. OSCRE’s Industry Data Model, an open access database, is one example of standardization, as are the performance standards suggested by LEED certification.
