Automated Valuation: What Zillow Offers Teaches About AI

Summary

Automated valuation looks simple: your client shows up with an estimate of their home, and you bring different comps and a different price. In this video, based on chapter 2 of The AI Real Estate Disruption, Tino Aguirre explains what the 2021 shutdown of Zillow Offers taught us.

The lesson isn’t that automated valuation is useless, but that estimating and buying are different things: a margin of error is fine in a ballpark figure, but when you buy, you pay for that error. The answer is to combine AI to analyze many properties fast with experience to see what isn’t in the data.

Key points: automated valuation

• An automated estimate and an agent’s comps can produce different prices.
• Zillow shut down Zillow Offers in 2021 because forecasting prices was more uncertain than expected.
• Estimating and buying are different: when you buy, you pay for the error.
• Two documents for the same property can show different square footage without anyone lying.
• The best combination is AI to analyze fast and experience to see what isn’t in the data.

Frequently asked questions

How accurate is automated home valuation?

It works as a ballpark, but it has a margin of error. That's fine for orientation; when you buy, you pay for the error.

What did the shutdown of Zillow Offers teach us?

That forecasting prices to buy and resell homes was more uncertain than expected. Estimating and buying are different things.

How should agents combine AI with experience?

Use AI to analyze many properties fast, and experience to catch what isn't in the data, such as discrepancies between documents.

Transcript

Your client shows up with an automated estimate of their home. You bring different comps and a different price. Who’s right?

I’m Tino Aguirre, AI CTO at TREM Group and IDXBoost. I build AI for real estate and use it every day with my team.

In 2021, Zillow shut down Zillow Offers, its business of buying and reselling homes. The company said forecasting prices turned out to be more uncertain than expected.

The lesson isn’t that automated valuation is useless. It’s that estimating and buying are different things. A margin of error is fine in a ballpark figure. When you buy, you pay for that error.

Two documents for the same apartment can show different square footage without anyone lying. Before calculating better, you need to know what you’re comparing.

My bet is to combine both: AI to analyze many properties fast, and experience to see what isn’t in the data.

If an AI is right on average but wrong on exactly the properties you work with, what would you need to measure before following it?

At IDXBoost we build AI tools so every valuation comes with its context, not just a number.

And if you’re a realtor, run a real estate team or own a brokerage, TREM Group is the all-in-one real estate marketing, technology and growth platform for your business.

And my full book is free at tinoaguirre.com/book.