James Miller discusses how search will evolve following Google’s decision to buy Pyra Labs, and how that could further impact advertising, because “a superior search engine could challenge both big media companies and advertisers. A media company’s most valuable asset is its brand name. Since you can’t read everything, you must make quick, often uninformed decisions about what to read. Brand names save time by signaling quality.” An example of a strong media brand is the New York Times.
Miller discusses an idea which we have been thinking of for BlogStreet.
Imagine using some idealized search engine that could determine which news articles you should read. If you trusted the engine you wouldn’t care if it presented an article published in the Times or an anonymous blog.
You don’t get any more pleasure, per se, from reading an article in the Times than our anonymous blog; it’s just that there is a higher probability of you liking the Times’s article. This advantage causes quality writers to want to publish in the Times, which further strengthens the Times’s brand name. If, however, some search engine found the articles you would like regardless of publication place, it would no longer matter to reader or writer where an article appeared.
Such an advanced engine could be based upon your rankings of previously read articles. The engine could find articles whose contents matched others you have liked. It could further use the rankings of readers who have preferences similar to yours to predict which articles you would most enjoy.
Amazon.com currently allows readers to rank books and recommends new books based upon such a system. Since we read far more articles than books, however, a futuristic news search engine would have more data to work with than Amazon does.
This search engine might first present an article to a few readers. If they ranked it highly, the article would then be shown to a few more. Articles that attracted consistently high rankings would earn large audiences. If numerous people participated in such a system, any one user would need read only a few articles that many others had not already ranked. An article’s success in attracting readers would be based upon its quality, not its place of publication.
When we visit a website, by following specific links we are making choices. This “trail” is only known to the website, which could looking at the trails of its visitors, suggest new content based on what others have read. But this new content would be only from its own website.
What blogs do is create a trail across all the websites. Bloggers leave the trail on their own websites (blogs) of what they like. Taken over a large number of bloggers, this can potentially work like a content recommendation engine, just as Amazon can analyse the click-trails and purchasing patterns for book lovers.
This is one of the challenges I’d like us to work on in BlogStreet. Bloggers as Information Ants (or Filters), each making their own local decisions, but as a collective, we can now for the first time start seeing patterns on the whole.
Google has done this very well by analysing the links for its Search technology. The time is now ripe for a similar engine for recommending Content, based on links analysed from blogs.