As I understand things, Zillow's approach wasn't a corner, but rather an attempt at arbitrage/value-creation. They were pretty sure they had an edge at price-estimation and had enough capital to be able to simultaneously offer sellers execution speed and offer buyers a price they would accept. It is a market-making play.
My impression here is that Zillow has re-evaluated some part of that calculus and decided to pull back. I hope that they'll be able to retool and revisit the approach after learning expensive lessons, as it is my impression that there should be real value here for all concerned.
I'm pretty sure that their internal dataset is far more rich/detailed than the public-facing "Zestimate".
Traffic-data alone, segmenting the buyers clicking on each listing and their expressions of intent, would give them an incredibly-actionable trading/training signal.
Zillow might see, nationally, the nationwide (and regional) direction of buyer and seller sentiment 2-4 weeks ahead of all but the largest regional brokerages.
My impression here is that Zillow has re-evaluated some part of that calculus and decided to pull back. I hope that they'll be able to retool and revisit the approach after learning expensive lessons, as it is my impression that there should be real value here for all concerned.