Contact Sales

Of all the booths in all the real estate conferences in all the world, she walks into mine.

I knew I was losing her the minute she sat down. I had seen that look before, the one that says, “I’ve been promised the sun and stars, so I’m not going to listen to anything else.”  But retailers like her don’t come along very often, and this was Vegas.  Everyone has that glazed look in Vegas.  And this was RECon, the biggest real estate related show in the world.  Full of vendors, investors, brokers, deal makers, peddlers, hucksters, snake-oil salesmen, and now even “Data Scientists”.

“Data Scientists!” The words roll off my tongue like a mush of ball-bearings in bananas.  You know what they say: “There’s lies, damn lies, and statistics!” Well, now the statistics have been Vegas-ized.  Full of glitz and glamour with terms like “Machine Learning”, “Artificial Intelligence”, and “Neural Networks”.  It’s all too much, especially after learning about “Juice”, “Plunger”, “Stickman”, and “Vigorish” while gambling away my last paycheck the night before.

The A.I. magic wand had been waved before her eyes, and it was up to me to bring things back down to earth.  If I didn’t, she was headed for heartbreak.  The heartbreak of empty promises and lofty expectations unmet.  I’ve heard this story a thousand times before:  real estate decision makers like her who had been led down the primrose path of comically accurate machines making decisions for you.

Maybe the situation today is more like “The Music Man” (you got trouble…right here in River City, and you need A.I. to solve it); or like Frankenstein (Stanford scientists create out-of-control technology buzzwords).

So why am I going off the rails here? Because as Pogo says, “We have met the enemy and he is us.” This confusing mish-mash of blinding terminology and empty promises is getting kicked around like a hacky sack at Bonnaroo, and we are the ones doing the kicking.  All of us are jumping head-first into a ball pit of trendy marketing terms.

It’s time to level-set.  What do these terms mean?  What technologies are relevant?  How can we use them in our real estate decision making process?

If you spend some time Googling these terms (and boy, have I) you will start to see a consensus (almost) about how they are interconnected.  Generally, Artificial Intelligence is the broadest term, and ideally refers to a system which can mimic human decision making for some specific task, with possibly even greater accuracy. Examples include image pattern recognition, speech recognition, and language translation.

One key element of this is “adaptivity”, the ability to improve performance by learning from experience.  This “adaptivity” is handled by Machine Learning techniques such as: Classification, Clustering, Linear and Logistic Regression, Decision Trees and Random Forests, Support Vector Machines and Adaptive Boosting.

Some of those may sound familiar if you took statistics in grad school, some not so much.  So, machine learning is just a rebranding of advanced statistics, right?  Not exactly.

The major difference between machine learning and statistics is their purpose. Machine learning models are designed to make the most accurate predictions possible. Statistical models are designed for inference about the relationships between variables.” (March Toward Data Science: Matthew Stewart, PhD)

Which takes us to the main point.  Do we really expect these models to make actual real estate decisions for us?  Is your MIT/Stanford A.I. going to drive to a market and select the actual parcel?  Of course not.  What we really want from our analytics are actionable insights.  Something that helps us understand the critical drivers that should be guiding our strategy.  A.I. can’t produce your strategy for you, but “analytics”, generally speaking, can help you develop it, and some machine learning techniques (aka, “statistics”) can play a part in that.

Don’t get me wrong. A.I has its place in the world.  You want to model the weather?  Conduct split-second arbitrage trades in commodities markets?  Develop a facial recognition application?  A.I. is your friend.  But what’s notable about all these examples is that they exist in scenarios with rapid feedback about success and failure.  The success of a real estate decision won’t be known for years.

Business decision making will always be a blend of art and science, and the science doesn’t help if it can’t provide meaningful insights.  In real estate, it’s not useful to model sales of existing stores down to the fourth decimal place.  What matters is to develop a meaningful understanding of performance drivers so that decision makers can develop a more effective strategy going forward.  A static model of today won’t really help when it’s built on top of algorithms and mathematics that decision makers can’t learn from.

Oh, and one more thing: data!  We haven’t even talked about the need for quality data to feed these models.  I guess that’s a topic for a different blog.

In the meantime, we’ll always have Vegas.