IIRC Zillow did "too well" to the point that they felt their model wasn't gauging the market correctly and was going to inevitably fail them spectacularly. The fact that it did well, better than it should have was just good luck... but also a bad sign because the model predicted a different outcome.