ClickZ, covering the launch of ThomasB2B.com (a search/directory service focused on the B2B vertical) has an interesting quote from Dan Savage, the president of ThomasB2B.com:
“I think if Google or Overture were to start over again to build the ad model that they have, they would model it around categories rather than keywords. They used keywords because they were there — people went to search sites and entered query strings. As bidding categories, bidding strings are quite clumsy,” he said.
Taking the opposite perspective, Google said the following in their release announcing an update to their Local Search offering today:
“This innovative approach does not rely on pre-determined categories and enables Google Local to provide local information that exactly matches a user’s search. For example, a woman in Palo Alto, Calif. could be looking for a restaurant but is on a budget. She would have trouble finding listings that provide this information in the yellow pages or on other local information sites. However, by simply searching for [cheap restaurants] in [Palo Alto] on Google Local, she would find a number of nearby restaurants to fit the bill. Google Local achieves this level of relevance by searching restaurant reviews and other related web pages, in addition to business listings, to return results described as [cheap] on the web.” (Ed: Emphasis mine.)
Mr. Savage is thinking like the President of a print company, not a search company. In print, you have to define a fixed category for businesses; if you listed every business under every permutation that a user might conceive of, your print offering would balloon to huge proportions. Instead, print products, like the Yellow Pages, solve this by heavy use of internal indexing and referral — e.g., “See Also” — for terms that user’s frequently use instead of the category that the YP company has seen fit to categorize the business under. For example, try looking up “pizza”; you’ll be referred to “restaurants->pizza”, even though 95% of people hitting their YP are looking for pizza the food, vs. pizza industrial equipment, etc.
Yes, keywords are a bit clumsy from an advertiser perspective. (And in fact, that represents an opportunity for your friendly neighborhood SEM, who would be happy to bundle up a set of recommended keywords for your business…) But unlike print, in web search, the advertiser must bear that clumsiness to make search quick and easy for end users. This will evolve eventually, as concept clustering, personalization, parsing of intent, etc. improve result sets, decreasing the effort for all involved parties.