The oft-repeated and now legendary story of a person asking an expert about the most important aspects to consider when buying real estate only to get the response “location, location, location” remains relevant but offers an incomplete picture of what is often a family’s biggest investment decision and what, in aggregate, is the world’s biggest class of assets- residential real estate. No doubt, location is incredibly important but so are scores of other factors. As both individuals and large financial institutions alike attempt to ascertain true values of both individual homes but also vast swaths of real estate, the imperfections and simplicity in the traditional ways of thinking about Real Estate values are brought into bold relief.
If the answer to these woes must be written in triplicate then I believe we need to change the “location, location, location” meme to “data, data, data.” We then need to ensure we add in that “data, data, data” is only as good as the Artificial Intelligence used to interpret and learn from the data in order to collapse on anything close to true value in this most fundamental of economic- and personal- categories.
At the surface, such tasks might appear easy. After all, isn’t all the necessary data public and isn’t “processing” data old hat by now? Such interpretations are understandable; after all, why would someone in another industry decide to get a Ph.D. in Real Estate. Further, with all the claims about Artificial Intelligence, isn’t that just “table stakes” in today’s world?
For veterans of the Real Estate and Technology industries, these questions lead to far more complex and imperfect answers. We find, in fact, that an incredibly large set of the relevant data required to understand even how to value real estate (not to mention that required to make analytical predictions about “imminent” or “future” states like propensity to list a house for sale, refinance, default, purchase a second house etc..) are at once NOT public and simultaneously far too raw to utilize for any real decision-making. Put simply, the data sets required to understand even valuation processes are vast, come from scores of disparate sources, and require a great deal of intelligence added to be useful for decision-making. With $22 Trillion at stake (just in the US) none of can believe that “good enough” is good enough!
AI and Machine Learning are incredibly salutary entrants to this mix. Given that the relevant data sets are at once huge, disparate and uncleansed (raw) but also constantly changing and that too in a non-linear fashion, it is impossible to create a valuation or propensity model with one-off, static data and analysis. We require both a great deal of applied intelligence but also algorithmic learning to hone, tweak, and improve models based on both methodological concepts and real-world data at once. As such, real AI is needed and not just claims of AI. AI and “Doing Big Data” are not synonymous. Indeed, the real world is messier and requires more sophistication.
There is more. Applying intelligence to small numbers of items is challenging but is usually tractable. Doing so at massive scale in truncated and shortening time-scales- that requires the Herculean tasks that AI and Automation allow for. In the Real Estate industry, these concepts- at least as far as valuations goes—are enshrined in the concept of “AVM” (Automated Valuation Models.)
The AVM mousetrap is a huge innovation. Getting the best mousetrap built is a different matter. That’s what Clement Ifrim and John Smintina, founders of Quantarium, realized when they stumbled upon residential real estate as a great application for AI.
Of course, great AI cannot be kept in a bottle forever. Like the genie who escapes, great AI can suffuse other industries and power other applications. Once out, it cannot be put back in.
The Quantarium AVM model, QVM, is just such a genie. Ranked as the best in the industry, QVM uses advanced AI and Machine Learning- drawn from multiple disciplines and principles of Physics particularly, to value large swaths of real estate. The scale is awesome- there are 105 Million residential parcels in the United States, each of which has its own data and its own story.
The combination of industry-leading data sets and our AI-driven approach offers much more as well, from analytics to insights, from propensity models, to decision-engines.
What we have is real. It’s thrilling to see the results when organizations come across our QVM and related solutions and apply them to their holdings and ever-changing business challenges and opportunities.