Folding Data #21

An Interesting Read: is data to blame for Zillow’s $245M Loss?

The news is officially out that Zillow is closing down its iBuying business, ending its foray into home flipping. Despite absolutely crushing it on name-brand recognition, traffic, and owning all the SEO keywords, Zillow simply couldn’t make a sustainable profit flipping houses during the tumultuous pandemic real-estate market - even as OpenDoor saw significant success in the same industry. Since the ability to correctly estimate and forecast a home value is essential to the thin-margin iBuying model, many speculated that it was a failure of Zillow's Data Science team who overpromised and underdelivered on the model accuracy, misleading the less sophisticated business leaders. The story below, however, presents a different point of view. What's most interesting to me is the sheer amount of leverage ($3B) that Zillow management bet on the accuracy of an ML model.

In Defense of Zillow’s Besieged Data Scientists

Tool of the Week: Supergrain 🌾

There's barely anything more frustrating than sitting in a room with two teams arguing about whose calculation of conversion (or any other metric) is the correct one. Launched publicly just a couple of weeks ago and led by Lyft's ex-Head of Analytics George Xing, Supergrain provides a solution to a growing problem of metric inconsistency and duplication of work that haunts every company as its data platform evolves. Supergrain's approach is bold in that it proposes not only a platform for storing metrics but also a framework to define metrics in a Lookml-style way and its own query language – SGQL, a SQL-like language optimized for querying and slicing metrics. The product is still young but I am very curious how it evolves and fits in the modern data stack!

Before You Go

As seen on Reddit.

‍