How Zillow’s Home-Buying Business Blew Up
Zillow blew up last week – shut down their home buying business division and laid off a quarter of their workforce.Why?Some data science warriors on Linked-in are waging a war on Zillow’s use or poor use of Data science and machine learning – specifically the Prophet library from Facebook, justifiably so. See recent posts by Peter Cotton, PhD and Valeriy M., MBA, CQF. Great explanations and references by both of them.Are incompetent data scientists to blame?Is Prophet to blame? From the examples Peter posts (as Microprediction), it appears prophet is an awful package – is that why it is free? Why facebook will provide something like this to the DS community? I am not seeing Facebook business blowing up – they are ready to conquer the Metaverse!!Is zillow management to blame?As the Barron’s article correctly points out, Zillow made some poor decisions buying in Euphoria when the market was too frothy, banking on even higher prices when the market slowed down.When you have debt and a low margin business, even a small slow down will hurt your bottom line.Finally no management would/should make calls dumbly based on models – regardless of whether it is prophetic or from Prophet.Stat/data science/ML/DL/probabilistic forecasting can only go so far. They are all wrong – we can debate how wrong all day.There are some natural phenomena that are more predicatble – but economic behavior not as much. Anyone that claims they have the science or analytics to predict near perfectly is a perfect con artist.Zillow management decided to double their inventory this summer – this must have been a management stretch on top of whatever result the prophetic models predicted.Not very different than what Solomon Brothers, Bear Stearns, Smith Barney or AIG did in 2008 betting on worthless real estate at the height of the bubble.