- DSNews - https://dsnews.com -

Man vs. The Machine: Big Data in Real Estate

This piece originally appeared in the April 2022 edition of DS News magazine, online now [1].

Big Data has been a hot topic in marketing and business circles for the past 10 years, and real estate is no exception. The industry is seeing an escalating push to leverage market and sales data to automate the buying and selling of homes.  

Notable industry leaders have dipped their toes into the water and developed automated valuation models (AVMs) intended to inform pricing decisions. However, in the face of a global pandemic and an historic shortage of homes, even the most robust models are struggling, with Zillow going so far as to shutter its iBuying service late last year. 

At its core, real estate is a people business, not an algorithm. Just like people, houses are all individual, with unique value propositions. It takes human insight to understand each and assess a home’s true value. While market data is incredibly important to support price decisioning, it’s not the “be-all and end-all.”  

The role of data and technology must supplement efficiency, not supplant or replace people. There are many dangers in focusing solely on Big Data.   

Big Data Requires Big Money
“Big data can help provide insights on thousands of comparable properties, instead of just a few, and can be harnessed to analyze market conditions and consumer profiles, among many other data sets, to establish property values more accurately,” explains Jayesh Maganlal [2], Chief Information Officer, DAMAC Properties. 

However, true Big Data requires extraordinary volumes of information from a wide variety of data sources, which is very expensive from both a technology and a talent standpoint. Real estate just doesn’t have the comprehensive arsenal of data available in other industries. Ours is limited in scope and transactional in nature.   

Artificial intelligence (AI) and machine learning (ML) routines require extraordinary input. If you don’t feed the routines with massive information, the algorithms will bias to the data they do have, which can be quite dangerous. Without enough data to be statistically significant, models may provide false automated valuations that do not represent actual market values in a fast-paced and dynamic environment like the one we’re currently experiencing. 

Big Data and Bias
“Research has long shown these computer models are loaded with biases and flaws [3],” said GeekWire co-founder John Cook [4]. Following Zillow’s move, the consensus among experts is that iBuying firms are putting too much faith in machines to do what humans can do better.    

Models usually aggregate and average MLS listing data at the MSA or ZIP code level. But with scant supply on the market, there are few comps to consider. And, as we all know, location, location, location is what matters in real estate.  

Pricing today happens at the micro-market, neighborhood, and block level, if not at the individual house level, with a range of aesthetic, social, and other factors playing a role, such as natural light, intuitive layout, finish level, etc.  

“The system may capture that there are three bedrooms, but does it capture that they are laid out in a way that makes sense?” comments NYU real estate dean, Sam Chandan [5]. 

Looking at suggested pricing based on city or town averages is highly inaccurate. You need qualified professionals with experience at the local level to assess pricing and offer customized expertise on localized buying patterns and preferences. 

A Backward View
Another important consideration in Big Data models is the timing of the data. By the time it gets to the Big Data engine, data is backward-looking—anywhere from 30 to 60 or 90 days old. That may be fine for a stabilized market, but not for a fast-moving market like the present. 

“All the AI and machine learning in the world isn’t yet up to the task of the complexity of valuing a home in a rapidly changing market,” notes MoxiWorks CEO York Baur [6].  

The pandemic has forever changed the role of the home where we now work, eat, play, teach, and learn. Space has become critical and layout needs are dramatically different. “That shift in buyer preferences is extremely hard for a machine-learning model to incorporate,” notes BiggerPockets data and analytics VP Dave Meyer [5]. 

Cue Zillow’s move to abandon the home-flipping business, a cautionary tale on the limits of Big Data. Its algorithms were just not able to account for the fluctuations in consumer needs and pricing that we have experienced over the past two years and to accurately predict future home prices and selling speed 

Consumers are tired of shelling out 6% to an agent who may or may not provide any real value in the transaction—and spending too much time and money on a traditional listing. While iBuyers want to give consumers an easier solution, the danger lies in relying on a computer to decide how much a home is worth. iBuyers may have experienced significant growth, but the question remains if that can continue in a market that is volatile and rapidly changing. 

The Right Stuff
So, what role should data play in a real estate decision? At New Western [7], we believe that data should help accelerate comparative market analyses (CMAs) but that local agents must serve as feet on the street to provide hyper-local intelligence and insight to buyers and sellers.  

Many experts agree with us: “What this says to me is that we need to stop over-applying technology in an effort to replace humans, and instead focus on applying technology to make humans better,” notes Baur [4]. 

We do see a place for Big Data and AI/ML in assessing opportunities for our agents.  These technologies can be combined in a way to find matches between seller opportunities and buyer preferences that result in the most likely candidate rising to the top of the list. Big Data has an important role to play in lead, opportunity, and deal scoring. 

Face the Future
A recent KPMG study [2] shows growing interest in digital transformation in real estate—from cost efficiencies to enhanced decision-making. Beyond pricing, data can be used to track demographic and employment trends and help developers identify and develop compelling properties. Add to that apps that use data to project potential income and earnings from a property.  

“A developer can thus quickly access hyperlocal community data, paired with land use data and market forecasts, and select the most relevant neighborhoods and type of buildings for development,” reports McKinsey & Company [8]. At New Western, we too leverage data science analytics to look at predictive indicators and assess opportunities for market expansion. But we still rely on people to dig deeper and make the final decision on where we'll expand next. 

Big Data can also play a role in how properties are marketed. Agents can use search engine and advertising data to refine and target relevant audiences. The sales process is yet another area where data can be used to create models measuring visitor interactions on competitor websites in addition to tracking interaction with advertising.  

Data can also be analyzed to evaluate buyer preferences and strength by looking at credit scores, mortgage pre-approvals, and other public records. 

While real estate is certainly ripe for disruption and applications like this, many variables must be factored for pricing models to become more effective. The challenge will be how to identify predictive indicators to assess markets—and the current market volatility is going to make that difficult for the time being.  

There is a huge role for technology and automation but it’s not the primary one. People will always come first. Information is one piece of the real estate puzzle. But it will always require humans to assess the data along with personal instinct, intuition, and experience to make educated business decisions 

Data is a means to the end…it’s not the end.