The GeoPhy AVM: From Theory to Practice
Obtaining an independent assessment of property value, whether it’s a broker-opinion-of-value (BOV) or a third-party valuation report, takes time. For a lender, this delays getting a quote out to a customer. For that same customer (i.e., the investor), it means losing valuable time to close a deal. The delay between ordering a valuation and receiving a report — often 3-4 weeks — clearly impedes a faster, more efficient underwriting process.
At a recent RICS conference in NY, Mark Snow, Chief Appraiser at Citigroup, discussed the time friction in the current appraisal process. He wondered out loud: “wouldn’t it be amazing if you could get a valuation by a click of the mouse.” Well, Mark, the wait is over. You can now indeed get a valuation online, via a click of the mouse (admittedly, a few clicks). We previously explained how the GeoPhy AVM works — through a combination of extensive locational data, tens of thousands of comps, and machine learning methods. These valuations are accurate, transparent, and instant.
Enter Evra, the GeoPhy platform that empowers the commercial real estate sector to access automated valuations, underlying value drivers, and detailed information on nearby amenities (akin to a ‘virtual neighborhood visit’) for multifamily assets.
Evra shifts the discussion of commercial real estate AVMs from theory to practice. All you need is the address of any multifamily property in the US (with plans to include other property types and regions on the roadmap) to get started. In the example below, we use a multifamily property in Portland, Oregon, financed by Freddie Mac, and securitized through the K-Series. Additional inputs required to run the valuation are straightforward:
• number of units,
• year of construction, and if applicable, year of renovation
• net operating income (NOI) for the property, based on the last 12 months. As an essential input for commercial real estate valuation, the Evra platform provides an NOI analysis and the discretion to adjust this input as needed.
Once you click ‘run valuation,’ Evra links the property address to its geocode (i.e., longitude and latitude attribution), which allows us to append location and market data to the property. In addition to amenities, such as schools and restaurants or local house price development from Zillow, Evra also integrates macro-economic data such as interest rates and current stock market conditions that strongly influence property values. The result, a rich resource of property characteristics and location information that feeds the GeoPhy AVM.
The GeoPhy AVM lives in the cloud and digests the data through an API. It takes just a few seconds for the model to run the valuation and deliver a fair market property value that includes a detailed set of underlying information. This ‘model assessed value’ is current, but it also provides a forward-looking perspective that includes transaction data that reflects future outlook and contextual data that is somewhat predictive of future trends. GeoPhy also has the ability to run retrospective valuations, but we currently run those valuations outside of Evra.
The model assessed value for 2813 SE Colt Drive in Portland, Oregon is US$88,606,000 — slightly below the previous property valuation dated December 2018 (in the current part of the real estate cycle, the GeoPhy AVM is typically less bullish than an appraiser).
The AVM derives the cap rate by simply dividing the NOI by the valuation. That’s a subtle, but important feature of the model. We don’t base our valuations on averaging cap rates of nearby, comparable properties. Instead, we estimate the property value directly based on a large dataset of property transactions, spanning the nation and two decades. While the model provides a highly accurate assessment of value, there remains some variation in the confidence level provided for each valuation.
The “confidence indicator” provides an indication of how similar the property is relative to all buildings in the GeoPhy dataset of property transactions, based on location, size, and vintage.
• Green. The property consistently aligns with observed properties, and we have a high level of confidence in the valuation — as is the case with the above Portland property.
• Yellow. The property may look a bit different from what we typically see (e.g., it may be a very small asset).
• Red. The property is not frequently observed, (e.g., the number of relevant transactions is low or ‘exemplary’ and thus not very comparable — a situation often experienced in New York City).
The confidence indicator is correlated with the expected precision of the valuation. In case of red, that means, treat with caution.
Of course, a valuation does not stop with providing just a number. In a traditional valuation report, you typically find at least 150 pages of information on location characteristics and (local) market conditions that may affect the value of the property. The GeoPhy AVM integrates that precise information, now quantified and translated into clear, objective data rather than long swaths of text. These so-called Value Drivers are provided in Evra through a selection of data on property characteristics and location and market details. In a machine learning model, relationships between property/location characteristics and value are not necessarily linear, so the interpretation of an additional instance of crime is not simply minus X percent or dollar in property value. Instead, Evra shows the relative importance of each property, location and market characteristic for the valuation of a particular property, including the direction of the relationship.
For the Portland property presented here, the density of the immediate neighborhood, for example, is relatively low — measured by the number of intersections in a 15 minute and 30 minute walking distance — resulting in a negative impact on property value, but there are many amenities within driving distance, resulting in a positive impact on property value. We also present many other location and market characteristics in Evra, including the absolute values for each feature (e.g. median rents for single-family homes in the area) and their relative impact.
The last part of each valuation shows an in-depth view of the immediate neighborhood — we call this a ‘neighborhood deep-dive.’ For both walking and driving REACH (i.e., the area accessible within 15 and 30mins walking or driving), Evra provides the count and location of educational facilities, restaurants, healthcare facilities, nightlife locations, retail facilities, and public transport access points sourced from Factual. So, rather than using Google — or that extensive 150+ page report — a real estate professional can now begin to understand the quality and vibrancy of a neighborhood through data.
And the current version of Evra is just the beginning. The Evra roadmap includes the inclusion of crime incidents for the next release, with a breadth of other data sets and functionalities — such as the addition of benchmarks to better understand locations relative to others nearby — in exploration.
For lenders, equity investors, CMBS investors, and bank regulators that frequently value large portfolios of property assets, Evra provides the option of batch processing, currently done through match-and-append. GeoPhy can also enable an API to provide direct access from our client’s systems into our Evra application. That means property portfolios can be valued, instantly, at any frequency required — be that yearly, daily or even intra-daily — with limited human interaction.
There is some element of futurism around the use of automated valuations in the real estate sector. But the future is here. AVMs are prevalent, even commonplace, in the residential real estate space. And for commercial real estate, real estate valuations can now be done using a combination of big data and machine learning models. Accessing those valuations no longer requires a spreadsheet, endless hours of desk research, or use of complicated worktools. With Evra, it simply requires “a click of the mouse.”
Try Evra and let us know what you think!