What does Computer Vision have to do with the Price of a House?

A guest post by: Romi Mahajan, CMO Quantarium

Residential real estate – peoples’ homes – is the world’s largest asset class, tipping the scales at almost $200 trillion worldwide.  This number is staggering to many, including those in the housing industry.  Larger even than the sums involved are the emotions – a family’s residence is likely its largest investment and one from which, so many other life-factors radiate:  Who are your neighbors, what schools do your kids attend, are you safe, how close are you to good medical care, and so on.  Insofar as this is true, the housing sector can never be given too much attention by Economists, Sociologists and even Technologists.  Still, in many ways the sector has been given short shrift.

Consider a matter at the heart of the industry – the value of a particular house.  What appears to be a simple question with a simple answer is not.  Sure, one can look at the basics – how big it is, the year built, comparable houses in the neighborhood and so on.  One can even attempt to factor in other variables – school district, crime statistics, proximity to the beach, and a host of other things.  All said and done, all of these factors are “external” and in many ways “non-specific.”

Let’s pause for a moment.  While these factors are indeed external, we have to ask ourselves a basic question – how do I get specific?  How do I assess the value of a particular house, looking beyond these basic factors and in the process taking into account the condition of the house and the nature of its interior landscape?

For most of us this is an obvious question.  After all, if you put in a lot of money to modernize or refurbish a house, you would expect that its value rises, even if your work and effort is not recognized in the external statistics being looked at for valuation.  If you, on the other hand, paid no attention to the house and allowed it to atrophy, you’d likely expect the value to diminish.

This issue is often “solved” by Appraisers, who theoretically take into consideration all of these interior and condition-based factors when assessing the value of a house.

Now, we enter a world fraught with problems.

For the purposes of this short piece, we won’t get into the debates about the objectivity of Appraisers or even about the shortage of talent that is delaying closings in many large markets (in the US for sure.)  These issues are fertile grounds for discussion, elsewhere.

No, the main issues we intend to dissect here are the issues of scale, speed, and customer experience.

In the US, there are over 100 million residential units.  Now imagine you work at a bank or other institution that originates and/or “owns” millions of mortgages and wants to determine the value of your portfolio in toto?  Imagine, further, that you need to do so every month.  After all, you need to keep track of your assets, make decisions about where to keep houses and where to sell houses, and assess your risk in holding these mortgages.  The issues of scale and cost are enormous.  You certainly can’t send an appraiser to each house.

Imagine a different scenario.  A consumer lives in a city with a very fast market and needs to make decisions on the spot whether to buy a house.  Waiting even a few hours, not to mention days, can mean losing a house.  In this cauldron, determining the true “value” of a home has to be done instantaneously.  Here, the issues of speed are paramount.

Finally, imagine you are a real estate agent with a demanding (and rightfully so) customer who wants to buy a house.  You have visited 10 houses to determine fit and have been disappointed by their dilapidated interiors.  You are not paid for your time, only for results.  If only, there were ways to determine condition and value based on condition in a way that was easy for the customer (in this case, you.)

Enter technology, specifically AI and its offshoot, Computer Vision.  Artificial Intelligence yields a potent set of tools for real estate, starting with valuations.  First of all, AI is “better at the basics” than non AI methodologies.  To get even a basic valuation of 100+ million properties every month is not trivial; with AI, the entire US footprint can be run in hours not weeks.  The idea is simple:  Computers can learn from data sets of a critical mass, then keep improving their outputs as more data comes in.  Machine-learning is just that- machines that actually “learn” and thus can offer results and outputs that are neither obvious nor simply the result of brute-force methods.  AI can thus help with the scale and speed components.

Computer Vision comes in here in a delightful way.  If you look at house-listings, they often come with a multitude of pictures.  Computer Vision can analyze and categorize these pictures- with speed and fidelity- thereby assigning “condition” scores to kitchens, bathrooms, and other hotspots in the house.  In this way, they can help offer a “condition-adjusted” value.

Put all of this together and you get a powerful mix.  Automated Valuation Models (AVMs), powered by AI can provide accurate valuations at scale and with enormous breadth.  Add condition-adjustment, powered by Computer Vision, and you start to see technology giving its due to the vexing problems and incredible opportunities in the real estate industry.

Written by Paul Dunay
Paul Dunay is an award-winning B2B marketing expert with more than 20 years’ success in generating demand and creating awareness for leading technology, consumer products, financial services and professional services organizations. Paul is the global vice president of marketing for Maxymiser a leading web optimization firm, and author of four “Dummies” books: Facebook Marketing for Dummies (Wiley 2009), Social Media and the Contact Center for Dummies (Wiley Custom Publishing 2010), Facebook Advertising for Dummies (Wiley 2010) and Facebook Marketing for Dummies 2nd Edition (Wiley 2011). His unique approach to marketing has led to recognition of Paul as a BtoB Magazine Top 25 B2B Marketer of the Year for 2010 and 2009 and winner of the DemandGen Award for Utilizing Marketing Automation to Fuel Corporate Growth in 2008. He is also a finalist for the last six years in a row in the Marketing Excellence Awards competition of the Information Technology Services Marketing Association (ITSMA), and is a 2010 and 2005 gold award winner in Driving Demand. Buzz Marketing for Technology, Paul’s blog, has been recognized as a Top 20 Marketing Blog for 2009 and 2008, a Top Blog to Watch for 2009 and 2008, and an Advertising Age Power 150 blog in the “Daily Ranking of Marketing Blogs.” Paul has shared his marketing thought leadership as a featured speaker for the American Marketing Association, BtoB Magazine, CMO Club, MarketingProfs, Marketing Sherpa, Marketing Executives Networking Group (MENG), and ITSMA. He has appeared on Fox News, and his articles have been featured in BusinessWeek, The New York Times, BtoB Magazine, MarketingProfs and MarketingSherpa. Paul holds an Executive Certificate in Strategy and Innovation from MIT’s Sloan School of Management and a bachelor’s degree in Marketing and Computer Science from Ithaca College.