The online get-a-price-for-my-house tools sure are a big hit. HouseValues.com has built a $100 million business in a few years essentially from this one consumer desire. Zillow, of course, made big waves this spring with a sexy UI and eliminating the hassle of agents calling. But it's the same real estate porn emotion driving that company's explosion onto the scene. After this long bubblicious ride, we all want to ask ourselves, "Am I a millionaire yet? What's Zillow say?"
The Wall Street Journal's little sister realestatejournal.com has an article following the various internet-home-price-estimator players and their widely varying results for a home in Seattle.
The trouble they point out is that we're all dabbling in a black art. To arrive at an estimate, you have to boil all of a property's characteristics into a mathematical/statistical framework. We've done this with one component of our market research report through a sophisticated pricing analysis of the on-market properties. You can start to zero in on an rough estimate price when you apply some machine-learning and natural language processing techniques to the property listing. But is the estimate ever good enough to bank your buying or selling decision on? Probably not. We like to think of the estimate of an potential-opportunity flag rather than a 'biddable' number.
As one of our Realtor clients puts it, "the estimates are all an attempt to commodify real estate. But it isn't a commodity." His point is that until you walk by a property and see, for example, that the layout is weird but the street is gorgeous can you arrive at a proper valuation in today's market. Because there are practically infinite variables to observe, there's no way to automate the analysis.
In the last few years we've seen the machine-learning and natural language processing techniques start to mature enough to potentially be applied to the vast variables of real estate pricing. What we don't have, however, is a good set of input data for those analyses. County tax and title company data is simple numeric stuff (square footage, beds, baths, lot size, age, etc.) Everything else is subjective. When we arrive at a way to categorically capture things like layout, decorative styles, relative appeal of the nearby shopping center, etc. etc., then the price estimators will be bankable.
So do the price-estimator-sites have any value? Well, if you want to know what your boss', or your in-laws' house is worth, roughly, knock yourself out. If you want to know where to price your house to sell quickly, get a good Realtor.