Last month I received an announcement about the hiring of a new CTO from the property data firm Cherre. Usually, I don’t cover new hires since I tend to get press outreach emails about new executives at companies almost every day. But this one got my attention. It turns out that Cherre’s new CTO has a rather prestigious background in data science. His name is Dr. Ron Bekkerman and he has a long career in academics and in tech companies big and small. He has been a professor at the University of Haifa in Israel, worked for companies like Google, HP and Motorola, was a founding member of the data science team at LinkedIn and has helped both startups and venture capital firms with their technical needs.
This hire is another in a recent spate of big-name hires for PropTech companies. I was curious to hear about his thoughts on bringing data science methodology to the property industry so I reached out for a chat.
The first takeaway I had from Dr. Bekkerman is that, even though there is a lot of real estate in the U.S., the size of the data isn’t really that big when compared to other industries, “The good news is that real estate in North America consists of a couple hundred million properties, which is an extremely manageable data set relative to many other industries,” he told me.
The goal that the team at Cherre is working towards is to develop a system that can provide precise information on every one of those properties, including who owns them, when they are being transacted on and for how much. “This data will allow us to systematically analyze exactly what is informing the decision behind transactions, and not only support individual transactions, but reliably predict coming trends in the market,” Ron said.
While the size of the datasets is not unmanagable by modern standards, the real estate industry does have a problem when it comes to uniformity. But it turns out that there is an answer for the incongruencies that come from contradictory data:
“One of Cherre’s primary value propositions is our ability to draw information from public, private and proprietary (often client-specific, internal) sources, and then resolve that data. We most often do that through the use of triangulation. For instance, there may be 10 sources of data on a specific asset, with nine in agreement and only one outlier. In that case, it is overwhelmingly likely that the nine sources in agreement are accurate. Over time, we create a weighted majority voting system, supported by our deep domain expertise, to incorporate the level of our trust in the accuracy of various data sources.”
Once he is able to complete that painstaking task of reconciling all the data he thinks that the benefit of creating a “truly predictive decision support system” will be worth the effort:
“As we build that comprehensive, clean data system for every real estate asset in North America (and eventually the world), we will increasingly be able to analyze exactly what is driving large property owners and underwriters, and what their strategies are. This is not an easy task, as ownership of many properties is often shrouded via small LLCs and other holding companies. Once a complete and wholly accurate data system is available, however, we are able to decipher the true portfolios of large investment groups and property owners and underwriters, and begin a deep analysis of the logic and overarching strategy behind their decisions.”
The ability to create predictive decision making is particularly important in the property industry due to the comparatively slow turnover of the products, “Real estate is unique in that its system of data objects does not undergo dramatic changes on a second-by-second, or even day-by-day basis. From a data science perspective, the buying and selling of properties is actually quite rare and slow-moving,” he said. The fact that much of the pricing data is stale in terms of statistical accuracy makes it even more important, in Ron’s mind, that a predictive tool like he is creating at Cherre exists.
There has always been a desire in real estate to have access to every important piece of data for every property in the country. The increased understanding that can come from this data set combined with advanced techniques like machine learning could be the next differentiator between the winning and losing in the property industry. With smart people like Dr. Bekkerman on the job, having this type of understanding seems like only a matter of time.