How about those Zillow estimates of price changes for next year? I guess now everyone knows they are garbage, since you can’t run a business based on them.
from the info I gathered, zillow has 2 models. One for zestimate and other for iBuying. Clearly iBuying failed mainly because of mgmnt buying homes more than the predicted value. Also I see some ppl say fbprophet was used in time series forecasting, so model was not great. This article says deep learning models were used for feature engineering. So it’s kinda confusing.
They never had a chance. They also had a mortgage business to pair with the home selling. The insanity was they only wanted to originate mortgages for the houses they were selling. They were terrified of competing against mortgage brokers and losing the ad revenue from them. Starting a mortgage business that’d be that small didn’t seem to make sense. It’d make more sense to partner with someone and collect a fee from referrals.