WSJ has more delicious bits From Seeking Alpha: So... 'cuz of this I read two thirds of Scott Galloway's piece of shit book. If you want to know what cloistered Manhattan liberal business professors thought of COVID in August of 2020, Scott's got ya covered. Why two thirds? He starts out with some interesting statistical arguments that between March and July 2020, business trends accelerated 10 years. Telemedicine? 10 years down the trendline. Online shopping? 10 years down the trendline. "Innovation?" 10 years down the trendline. Problem is? He presumes that there were no exogenous shocks to the system from a global fucking pandemic. He whips out that "weeks within decades" saw that Lenin never said and then ignores that whole October Revolution thing the quote refers to. So all his extrapolations are basically "invest as if it's ten years in the future and absolutely nothing has happened that is going to cause anyone to reconsider anything anywhere ever." Traditional economics assumes a certain inelasticity of demand. A constancy of want. Galloway goes as far as surmising that any couple spending March-July together experienced ten years of mileage on their relationship, for better or worse, and any employee ten years of employment... and gets as far as recognizing that's make-or-break on every level... but then doesn't even acknowledge that ten years of shock might just break his model. So Zillow had a model that they thought gave them an edge over the wankers on their marketplace. This is very much like Amazon using Amazon data to make better Amazon products to compete against the stuff they sell from other vendors, but whatever. Fundamentally? I get 200pp of graphs every morning. One of the greatest insights it has given me is that there's leading indicators and there's trailing indicators. Nobody ever breaks this out. It's just data in every analysis. But if you look at it, half of it is "here's what we saw" and the other half is "here's what we're seeing." If you can't tell the difference, your "here's what we saw" model will be used to guess what you're going to see. And if things are steady-state, it might even work for a while. Steady state is on hiatus.“It feels like this would be a hard time to lose money buying and selling houses,” said Benjamin Keys, professor of real estate at the Wharton School of the University of Pennsylvania. “This is a time frame where prices have gone up in a lot of places, dramatically.”
“We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated and continuing to scale Zillow Offers would result in too much earnings and balance-sheet volatility,” Mr. Barton said.
The move represents a big hit to Zillow’s top line. Home-flipping was the company’s largest source of revenue, but it has never turned a profit.
"We went into the business as a big swing on the bet that we could accurately predict the price of a home six months into the future, and what happened was... COVID happened," he later told CNBC.
and what happened was... COVID happened," he later told CNBC.
I'd be really curious as to how they measure the accuracy of the models they've been using, and then what the accuracy was measured to be. Haven't connected what's happening with Zillows iBuyer program to the housing market at large, maybe there's no real connection there outside of shit's crazy and still not a buyers market.
The thing about the Zestimate™? It was always a hype tool. It was always a "look how rich you'd get if you sold your house." It was always a "hey look what a great investor you are by beating this price." It was an armchair piker's metric, a bullshit statistic for the window shoppers that allowed random fuckheads to walk around thinking they knew real estate. One of my wife's college friends came down to LA for my wife's baby shower. We couldn't walk around anywhere without her checking to see what Zillow thought the neighborhood was worth. I could say things like "that price is utterly disconnected from reality and based on nothing, that house over there sold for $100k less than Zillow thinks it's worth" and my on-the-ground observations counted for nothing because an algorithm gave her Truth. Whaddayathink - did Zillow come up with a separate Zestimate™ in order to assess the revenue extraction from their predatory flipping practice? Or did they just get high on their own supply? That's Zillow, drinking their own kool-aid and then puking it back up. Because their model didn't say "uhhh you can't predict what the fuck's going to happen in the time it'll take you to buy and sell a house" their model said "two points determine a line" and they doubled fucking down.During the third quarter, Zillow said it bought 9,680 homes and sold 3,032 of them, with the sales producing an average loss in gross terms of more than $80,000 per house. The purchase of nearly 10,000 houses marked a significant jump from the 3,805 homes Zillow purchased in the second quarter, which itself was a record amount by more than 1,500 houses.