compstak and baya

The Path Towards Commercial Real Estate’s First Automated Valuation Model

“What is this property worth?” This has always been the most important question in real estate. Fortunes have been made and lost because of incorrect answers to this very query. For residential real estate this question has been largely solved. Appraisers can look for similar houses in the same area that have been sold in the last few months and infer the price. But commercial property is a whole other enchilada, one with way more layers of toppings to pull back to see the meaty, truthful filling (reminder to myself to stop writing when I am hungry).

Commercial buildings don’t change hands that often it can be hard to use sales data. Plus, most commercial properties are just that, commercial and therefore valued not by own much the market likes it but by how much income it can generate. So we use the revenue stream, or rent roll, and apply a capitalization rate similar to how company’s stock value is compared to its revenue. So the first step is to find out what a building has historically been rented for.

This is tricky because unlike sales information, there is no public record of this. So, a few technology startups have found ways to gather this data. One is CompStak. For the last few years they have built a healthy data set by sourcing lease comps from brokers and appraisers in return for generalized data. Another company to do this is flexible office provider Knotel with their new blockchain initiative Baya. They hope to similarly incentive property companies to input data but plan to do it by issuing coins that (ideally) will go up in value as more data becomes available.

Even though these two companies have somewhat competing offerings, they both see value in having the ability to compile enough data to be able to automatically value properties. So when Knotel acquired 42Floors, a marketplace for office space that has been around since 2011 (making it ancient in the PropTech space), Amol Sarva (Knotel cofounder) and Michael Mandel (CompStak cofounder) both saw an opportunity.

Mandel spoke in detail about his decision and plans in the PropTech Connect Slack channel, “So this deal is a proof of concept, and we’re starting with just NYC office. The big challenge with an AVM is that to be accurate you need the rental information for EVERY lease in the building, and there could be leases going back 20 plus years.” He continued, “Our database in NYC does go back 20+ years, but not with as substantial coverage as our recent data. The opportunity here is to use 42Floors data to help fill in the blanks.”

As Michael said, creating an automatic valuation model requires incredibly complete data. This means that they have to do some inference for information that they don’t know. Michael told the channel that, “For as far back as 42Floors data goes—something like 7 plus years—we can see when a space was on the market. That way if we know a space was on the market, but we don’t have the comp for that space, we can use an algorithm to imply what the rent would have been for that space when the lease was signed, by taking into account the rent on deals signed at that time in similar spaces and similar buildings.”

Michael also told me personally that this partnership is only for New York City office market and does not include data sharing. CompStak data will not be imported into the Baya database or visa versa and “there will be no coins involved.” This is a great example of how even companies with somewhat competing projects can collaborate to the benefit of us all. One day soon we might all be able to much more easily answer real estate’s eternally important question, “How much is this property worth?”

Propmodo is a global multimedia effort to explore how emerging technologies affect our built environment.

More Stories
The Opportunities and Challenges of Self-Serve Information in Real Estate