I think one’s perspective on the Zestimate’s “within 5% of median selling price” claim depends on how impressive or not you think the tool is.

One's perspective is more easily shaped by how much of a lie you think that claim is.

Here's a friend's house. Amazingly, it's been through three sales since they got out of it but hey, that's Playa Del Rey. Even more impressive are that the pictures are from 2015.

Zillow: $514k, despite the fact that it sold less than eight weeks ago for $508k. That's within 1% - but that's the tool AFTER THE FACT. Or is it?

Remax: $530k.

Realtor.com: $568k. Realtor.com smokes crack.

Redfin?

Redfin puts it at $523k. Redfin also lists sales, but doesn't list the last two - which explains why the pictures are from 2015. Realtor.com lists those same sales. Normally you can look this stuff up on the county assessor's site but Los Angeles County charges for access because of course they do. but there's every evidence that Zillow is straight-up making up sales. Not only that, they know you're a buyer, not a seller (because your zip code is nowhere near where you're looking) so it's lowballing. When I look up my own house Zillow is suddenly $10k higher than Redfin.

In my eyes, it’s kind of a miracle, especially considering how difficult it is to compare any two houses even if they’re in the same neighborhood and similar square footage, features, etc.

In my eyes, there are 50-odd houses within a quarter mile of me that are one of three plans. There are record of every equity-backed home improvement loan for each one of them. There are publicly available sales records and there are publicly available assessments (which, around here, come in at less than half what the real estate sites estimate for sales). Redfin and Zillow differ by $10k on how much my house is worth - that's a kitchen remodel. And it doesn't really matter whether I've replaced my storm windows or polished my hardwood - my price is set by other similar properties and in that aggregate, there's a normal distribution.

It's just data. The difference is, Zillow manipulates theirs.

Furthermore, the Zestimate has to always somewhat trail selling prices, due to changes in preferences that even experienced realtors have difficulty predicting, let alone excel spreadsheets.

This is simply not true. The Zestimate will trail selling prices to buyers and lead selling prices to sellers because they want to get people selling stuff. Also mentioned in the article is the fact that 2/3rds of Zillow's business is advertising to real estate agents - they're not working for you. They're working for the people trying to buy or sell your home for you.

But I think the plaintiff is partly right—everyone uses Zillow to gauge prices.

It has been my experience that people use Zillow to gauge prices in markets where there is only Zillow. Anyone who looks at it for any amount of time determines that they're liars and crack smokers. I'm keenly interested about shit like sales-to-list ratios - that's what indicated the market was cresting last time. Zillow actively fabricates that data.