Fixing Inaccurate Zestimates: Housing Values Based on Machine Learning

This week in Tech Thursday we’ll go over a topic I have spent many years studying; Machine Learning. We’ll also cover how this applies to real estate.

Zillow Zestimates are well known to be inaccurate. Even though this trend has improved in recent years, there is some inaccuracy in these tools. This will soon change thanks to Machine Learning.

What is Machine Learning? I’m glad you asked. This is a big year for the term and it’s a little blown out of scale. Border-line buzz-word, Machine Learning refers to predictive algorithms, aka math. With enough data, statistical models can fit nearly any problem. Think of it as a combination of statistics and time. For a simple example, an average is generally not going to allow you to predict every element of a set. Yet, if you refine your formula enough, you can end up with a function that fits nearly every data point. This can then allow categorization on elements not in the original set.

The most popular Machine Learning technique at the moment is Deep Learning, or Artificial Neural Networks. This is the same mechanism that Google’s search engines use. It is based on how human brains function. A network of neurons processes a problem. After running through each neuron, the best answer is selected.

What does all this mean for Real Estate? It means that these tools like Zestimates are going to get better. Say good buy to those eye-popping inaccurate Zestimates! No more crazy numbers from sellers, either. The price a home sells for is more complicated than where it is. It’s a mix of many elements that an average number won’t entail. But a Deep Learning model trained with millions of homes sold can, and that’s the difference.

There’s also an element of prediction in place. referring to the ability to know when a home-owner might be about to sell. Just as Google search history can tell if someone is pregnant, it can apply to homeowners. There have been enough searches by home-owners thinking of selling that the indicators are there.

All-in-all, the entire home selling process is going to be shorter. It should also become more painless, for seller, buyer, and agent. The more expectations are set and met all around, the less discomfort there will be.