After hearing all of the hype about the new website,, of course we had to challenge it.

We gave Zillow a lot of leeway, deferring to its Beta status and its disclaimer that it was not up to speed in many localities. And it indeed it was not. There was no information about our home county in Southeast Georgia, nor in several others in semi-rural areas close to Atlanta. There were gaps in places where we thought that coverage would be better. A correspondent in Houston, where Zillow ranks itself with two stars and says it has data on most houses reported that neither the home she just sold nor the one she just bought appeared at all nor did her son's home in Austin. Another friend's home on a remote island in Puget Sound did better. Her entire neighborhood was on line but with only very limited information available from assessor's records. In Denver, rated by Zillow as having four stars worth of data, we found a lot of information although we were lacking the knowledge to evaluate it, and in Virginia there was a very strange blip which we will talk about later.

In order to give Zillow a fair shot we picked a town that Zillow advertises as having thorough coverage at present and is one which we know well. The town, in the Boston area, is where this writer lived for 32 years and sold real estate for nearly a decade. The town is very upscale and values there have increased at a rate that has far outstripped those in most parts of the country.

We were impressed by the information that was available from Zillow for this and a couple of surrounding towns.

We focused on a house, a nice cape in a fairly good location, and used it to put Zillow through its paces, both to illustrate its features and to evaluate the accuracy of its information. The house is a classic cape with a shed dormer, hardwood floors, and a medium sized yard. It has been owned by the same family since August, 1978 when they bought it for $65,000 in a very tight market. At that time it had a recent family room addition with a wood stove and new carpet and an averaged-sized state-of-the-art custom kitchen with hand crafted cherry cabinets and all of the appropriate bells and whistles at that place in time.

In a chance encounter with the owner in 2002 we were told that the house had changed little and, in fact, suffered from deferred maintenance - and that the only real improvement had been a very small deck off of the back door, constructed some years earlier.

It seemed perfect and we ran all of Zillow's capabilities using this house as our prototype.

Where information is fully available Zillow offers a satellite image of the neighborhood with every house tagged with a price. Where little or no data exists you might get a satellite view with no tags or only a dead-end tag with an assessment, or simply a street sketch. Our test case gave us the hybrid satellite view, all streets correctly labeled, and each house tagged. We found that we could not obtain any other information on houses in the view by clicking on the tab but clicking on the subject house gave us the following:

House description. As one would expect, this information is only as good as the contents of the assessor's field card. The information on our test house - bedrooms and baths, features such as fireplaces, lot size were all correct and the target Zestimate, $602,000, was not far from what I would have expected for the house and neighborhood.

Home Value range. This, as stated in the previous article, reflects the relative certainty of the underlying data; the more information that is available for Zillow to crunch, the smaller the data range. In this case the home value range was $535,727 to $686,212, a range of 22 percent. We thought this wide range might reflect the lack of recent sales data, but checking two other homes in the area, one of which sold in 1997 and another in August of last year we found that the ranges were equally wide. The house that sold last year for example, had an actual sale price of $600,000, a Zestimate of $646.20 (quite an appreciation) and a range of $575,136 to 736,691, again 22 percent.

Historical price tracker. This is a graph that allows the reader to track price increases for the house, the zip code, the town, county, state, and the country on a 1year, 5 year, and 10 year basis. The resulting information can be presented in a format reflecting total percentage, annualized percentage, or a total dollar figure for each period. In the case of our house, the historical trend has been 12 percent in one year, 51.1 percent over five years, and 132 percent for the last ten years. The average yearly increase was 8.8 percent. When we ran the annualized increase against a solid sales figure in 1978 we came up with a figure just slightly higher than the upper end of the value range.

Also available is information on how a house ranks among others in its Zip Code (ours was valued higher than 24 percent of its neighbors) town, county, state, and country. Our little cape placed higher than 85 percent of those in the state, but only 86 percent higher than houses in the country at large. Both of these estimates seemed off. Given other information we have seen we would have expected the house to rank much lower in state comparisons and much higher in national rankings.

Next we tried out the customization feature. While the science behind this is pretty simple and is just a formula based on information from Remodeling Magazine (November 2005 edition.), it is a lot of fun to play around with.

Using our test house we gave it a mid-range kitchen update defined as refinishing existing cabinets, new paint or wallpaper, and laminate countertops, etc. We valued our upgrade at $10,000 and said it was done last year. Zillow came back with a house value of $611,273, only slightly discounting our improvements. A second undefined improvement done within the last year and valued by us at $1,500 earned another increase of $1,000 in value.

Then there were comparable sales. When we requested these we got 10 recent sales, all within the year, most within the six months that most appraisers use, and all within a mile of our subject house. There was an equal amount of information on each house as on our subject and one could adjust these as well to reflect how improvements similar to those in the subject house would impact the comparables. At one point we also got comparables on 50 houses within the Zip Code. We couldn't figure out how this happened nor could we replicate it. This was an occasional problem either with us or with the web site when looking for or testing information.

Zillow conducted itself very well in our one-house test. We did, however, find some strange things about the service as we looked at it from several other directions. We will talk about those next and also evaluate another new site which hasn't received nearly the publicity of Zillow but is certainly receiving a lot of quiet hype.