Local search is getting a lot better, but one of the things that most engines and sites cannot now do with accuracy is deliver results consistent with the way people actually search. For example, I live in an area of Oakland, California, known as "Montclair." Montclair isn't an official place, and so it doesn't often show up in most engines (Yahoo! and Google sometimes get it right). If I use my ZIP, the engines have an easier time.
But people search for things according to neighborhoods and colloquial names. Think about New York. A local search in New York might use phrases such as "Upper East Side" or "SoHo" or "Midtown." It's equally true in London or Paris — the names used for places by ordinary people are not the DMAs or necessarily consistent with official municipalities or formal geographic boundaries.
So it's significant that real estate vertical HomeGain (owned now by Classified Ventures) has developed what it is referring to informally as "neighborhood maps." I spoke to Dan Martin at HomeGain recently, and he explained that the company had "mapped" 200 cities with more than 11,000 neighborhoods (I believe this is now U.S. only).
This is what Dan said in an e-mail to me:
We can now enable local searches for things like "Russian Hill apartments" or "sushi in Bernal Heights" [San Francisco neighborhoods]. To date nobody else we've encountered has developed good neighborhood maps on a national level. I think this could have immense value to folks in travel (think hotels), local search, real estate, yellow pages, etc. Analyzing data at the neighborhood level may also reveal insights that are not apparent when using zip code boundaries. For instance, Starbucks may want to put its shops in the center of a neighborhood or at the neighborhood junction points.
HomeGain is clearly going to use this, but Dan also said he is interested in licensing this data/technology to others.