Improving the Zestimate: An Experiment
Dec 11, 2018Update 11/11/21: This article was written originally in December of 2018 and, given the recent shutdown of Zillow’s home buying efforts, I felt it was important and relevant to revisit.
The team at ForxyAI conducted this experiment, which resulted in improving the accuracy of predicted home values in 60% of the homes tested, to highlight the importance and value of incorporating the current condition of the home into an Automated Valuation Model (AVM). After all, how can you possibly value a home without knowing if it’s been well maintained or if it’s a complete disaster?
In fact, the first sentence under “Disadvantages” on the Wikipedia page for AVM says, “The disadvantages are that they do not take into account the property condition, as a physical inspection of the property does not occur and therefore the valuation produced assumes an average condition which may not reflect current reality.”
This is why we started FoxyAI. To unlock the condition of a home using AI and computer vision. It’s imperative that when photos are available, the property condition is included in the valuation model, and the only way to do this at scale, across millions of properties, is to use AI.
Whether you use a 3rd party AVM or a custom in-house model, contact us to learn more about how our computer vision can improve the accuracy of your valuations.
FoxyAI set out to improve residential real estate valuations by incorporating image data. To do this, FoxyAI research developed FoxyNet, the Convolutional Neural Network that powers house2vec. House2vec takes a raw image and returns an image feature vector, embedded in a high dimensional space, containing information on the quality and condition of the property for use in valuation models, among many other applications.
We decided to pair house2vec with arguably the most famous and divisive state of the art valuation model, Zillow’s Zestimate. The results of several experiments will be discussed in a multi-part series.