Dennis Cisterna is the chief revenue officer of Investability Real Estate, Inc. Investability is an online single-family investment marketplace that utilizes the industry’s best data to provide cash flow, net yield, gross yield, and investment analysis on over one million properties for sale in the United States. DS News recently spoke with Cisterna about the increasing role of data and analytics in predicting trends for the mortgage market.
On the topic of using data to predict trends, is there a particular data set that you are using more often to forecast trends in the market for 2017?
Our whole algorithm is based on a host of different data sets. When you talk about how data science and business intelligence is involved when it comes to big data, it's because we're able to pull data in from so many different sources and then scrub it to take out the abnormalities and then put modeling on top of it—that's what's really helping data and analysis get better and better as it goes on. As our business has evolved, we have become much more savvy with understanding what are critical factors and what aren't in determining key components of our data sets. One of the things we're working on right now is literally a property score for every property on the scale of 1 to 100 in terms of its overall quality relative to its potential rent. It takes into account dozens of different factors and weights them based on their overall impact in that market. For example, let's say a property is rated an 80 in Oklahoma City. I'm sure that property is nice and the schools are good, but it's not going to look like a property that's rated an 80 in Los Angeles or Denver, because there are different dynamics involved that drive quality of life and value in each of those unique markets. This is something I could not even fathom us doing three years ago.
How has the role of data and analytics in the mortgage industry changed over the last 10 years? Computers and the Internet were in widespread use at that time. What is different now?
It's really a combination of two things: Number one, the data sets have increasingly gotten larger for almost every kind of industry or set of information you're looking at, and two, the processing speed of that information has ramped up because of the computing power and the ability for cloud storage. All these factors mean basically now we can access and analyze more data in a quicker fashion. Information 101 says the more data points you have, the more confidence you can have in that decision. Now we're at a point, and it's continuing to get stronger and stronger, where we are analyzing billions of pieces of data in a handful of seconds. That allows us, as a data company, to be able to convey more powerful information to the users that need it, whereas we didn't even have that as a delivery mechanism before. Ten years ago, if you even tried to work with big data sets, it would take a huge multiple of what it takes today, and you still wouldn't even have access to be able to synthesize all that information together the way you do now. It is light years in a decade for sure.
Now technology has improved so much in just a few short years that we’re analyzing billions of pieces of data in seconds now, where is there room to improve from here?
At that point, now that we can analyze large data sets, the next steps are detecting the signals from the noise. What I mean by that is, figuring out what variables are truly important for predictive modeling. Within the field of data science, machine learning has evolved dramatically to the point where we can detect trends in data with a higher degree of confidence than ever before. Real estate, by and large, is sometimes considered an old and stodgy industry that seems to be a little behind the times and part of that has been due to the high level of regulation, but the industry is starting to play catch up in a major way. The driver of all of this, of course, has to do with analyzing risk. This is especially true for lenders. We're getting to a point now where we have this continuing ability to access and analyze this overwhelming amount of information that wasn't even available before.Now we are evolving to that point where as these machines get smarter and smarter, for lack of a better term, they are running every scenario possible, and the more information you give them, the more that lowers the probability of their decision being wrong.
As an example, Sabermetrics and data science have taken over baseball to an alarming degree, so much so that teams trust a computer program more than a guy watching with his own eyes. I don't know if that means that business intelligence and machine learning and everything else moves us to the point where we start trusting our AVMs more than our actual appraisers, but it's certainly a very powerful tool now and it's made the quality gap much less. If you ask even five years ago about the difference between an AVM and an appraiser, most people would say, “It's not even close. I don't trust an AVM. That's just a starting point.” Now, you've had some serious advances in a number of tools where the difference is negligible. If you look at some of the groups that are doing these “next gen” AVMs when it comes to property valuations, I'm pretty impressed. I'm having a growing confidence in the right kind of AVM being a good substitute for a BPO and that is just the tip of the iceberg of what is in store for the real estate industry.