1. 1999: 37% of buyers searched for a home online. 2009: 90% of buyers searched for a home online.
2. 1999: median home value is $137,600. 2009: median home value is $172,600
3. 1999: 82% of buyers purchased detached, single family homes. 2009: 78% of buyers purchased detached, single family homes.
4. 1999: 46% of buyers choose suburban neighborhoods. 2009: 54% of buyers choose suburban neighborhoods.
5. 1999: 68% of buyers were married couples. 2009: 60% of buyers are married couples.
6. 1999 and 2009: the median age for buyers was 39.
7. 1999 and 2009: “neighborhood quality, affordability, and convenience to work and school have consistently been top priorities.”

I bought my Montclair home just over a decade ago –11 years to be exact– and it has doubled in value, current market conditions notwithstanding.  Happy to be living here in Essex County, NJ and not where the median home value has increased by a mere 25% (see #2, above).

(Lani Rosales, December 31, 2009, Agent Genius Blog)

Here’s the beef.  Zillow is a huge aggregator of data.  They use info uploaded by real estate agents (not always accurate), tax assessment data and comparable sales.  Then use a proprietary formula to calculate a “Zestimate.”  But they have no way of accounting for hyper local conditions, for example, a house that has not been updated for 50 years. Or a house that backs up to a gas station.  Or a house with a fabulous view (that the house next door may not have because it’s blocked by trees).  These are factors that only a human being with local expertise can work in to the house value equation. In fact Zillow’s self-reported level of accuracy in the Northern New Jersey Market is a median error of about 12 %.

In other words, half the houses sell for within 12% of its Zestimate.  And the other half do not.

Only 25% of homes sold for within 5% of the Zillow estimate.  That means you can count on Zillow to be really accurate only about a quarter of the time!

When Zillow is wrong, it can be really wrong. Take this house in Glen Ridge which recently sold for 380,000.  Zillow’s estimate was 573,000.  Can you imagine the agent trying to convince the seller that despite what Zillow says the house is actually worth almost 200,000 less?  Here’s what Zillow didn’t know about this house.  It needed extensive repair and was owned by a bank that wanted to get rid of it as fast as possible.

The take away is that no computer can substitute for a careful assessment by a qualified human.  While Zillow may work well for communities that have a very homogeneous stock of houses (think Levittown), it does not work as well for towns like Montclair and Glen Ridge where a two million dollar home is often a block or two away from a four hundred thousand dollar home.

On the bright side—isn’t it good to know that human experience and judgment still beat out the computer?