From the wsj.com:

Jason Gonsalves worked hard to turn his 6,500-square-foot stucco-and-stone home in the suburbs of Sacramento into the ultimate grown-up party pad, complete with game room, custom wine cellar and an infinity-edge pool overlooking Folsom Lake. When interest rates fell recently, Mr. Gonsalves, who runs a lobbying firm, looked into refinancing his $750,000 mortgage. That’s when he got startling news—the home had dropped more than $200,000 in value while he was renovating.

Or at least, that’s what one real-estate website told him. Another valued the house at only $640,500. And these online estimates left him all the more confused when a real-life appraiser, assessing the house for the refinancing loan, pinned its value at $1.5 million. “I have no idea how those numbers could be so different,” Mr. Gonsalves says.

Right or wrong, they’re the numbers millions of consumers are clamoring for. After years of real-estate pros holding all the informational cards in the home-sale game, Web-driven companies like Zillow, Homes.com and Realtor.com are reshuffling the deck, giving home shoppers and owners estimates of what almost any home is worth. People have flocked to the data in startling numbers: Together, four of the biggest sites that offer home-value estimates get 100 million visits a month, with web surfers using them to determine what to ask or bid for a home, or whether to refinance.

But for figures that can carry such weight, critics say, the estimates can be far rougher than most people realize. Valuations that are 20% or even 50% higher or lower than a property’s eventual sale price are not uncommon, as the sites themselves acknowledge. The estimates frequently change, too—sometimes by hundreds of thousands of dollars—as sites plug new data into their algorithms.

Appraisers and real-estate consultants say the online models can veer off target with alarming frequency. Most data for the models come from two sources: records from tax assessors and listing data for recent sales. Collection is a challenge, however, because not every county tracks properties the same way—some calculate home size by number of bedrooms, others by overall square footage. And automated models aren’t designed to account for the unique construction details that often make or break a deal, or for intangible factors like a neighborhood’s gentrification. “You cannot use a computer model in certain areas and expect the value to come out right,” says John May, the former assessor of Jefferson County, Ky., which includes the state’s largest city, Louisville.

For all these reasons, models that banks use often add a “confidence score” to their estimates. Consumer-oriented sites, meanwhile, rely on disclaimers, some of which are eye-opening. Zillow surfers who read the “About Zestimates” page find out that the site’s overall error rate—the amount its estimates vary from a homes’ actual value—is 8.5%, and that about one-fourth of the estimates are at least 20% off the eventual sale price.

The sites argue that, over time, edits and corrections will help them perfect their numbers—with many fixes coming from their customers.

Zillow has accepted revisions on 25 million homes—perhaps the strongest testament to how seriously consumers take its estimates. Today, the site says its figures are accurate enough to give consumers a good sense of any home’s value. In the meantime, says Mr. Humphries, its economist, “We’re always tweaking the algorithm or building a new one.”

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Automated-valuation websites to check out:

www.zillow.com

www.cyberhomes.com – go to ‘Home Values’ page and input the address

www.homes.com – go to ‘Home Values’ page and input the address

https://www.chase.com/online/Home-Lending/home-value-estimator.htm