Good article in the WSJ this morning (username/pw required) about the growing importance of personalization in search. The approaches discussed:

  • Google - Statistical tracking of web and search history to deliver more personalized results. A search for ‘Giants’ in NYC would return a football team whereas in SF it would return a baseball team.
  • Yahoo - Social Search based on user reported preferences around tagged sites, friend networks etc. Yahoo’s experience suggested that statistical personalization didn’t create enough variance in the search results, according to Eckhart Walther, VP Products, Web Search.
  • Pre-found - Human-indexed web search based on user submitted links. In addition, user submitted profiles which then are used to customize search results.

Some of the pitfalls of personalization that are discussed are obviously privacy but also the negative user experience created by getting it wrong. Another issue not mentioned, but that I think is salient is that of discovery. Personalization can’t just be based on what you have done before.

Software shouldn’t require users to actively report their preferences in order to be smart about personalization. It should be possible to track things like clicks, page views, time on page etc and couple that to the content in question to deliver a more reader-tuned experience. If a user does enter a preference somewhere in the system, you should take advantage of that, but requiring input up front doesn’t make sense to me. We have to deliver value to users before asking them for anything.

Source: Search Engines Seek to Get Inside your Head by Jessica Vascellaro and Kevin Delaney