caveman - 7:57 pm on Mar 26, 2006 (gmt 0)

Personally, I think that this thread shines light on the essence of the problem with hand coded SERP's. I understand the temptation to go down that road, but IMHO, it's a very poor decision for any search engine to make if the SE aspires to be a top quality player.

We all know the issues: The nature and structure of different categories (i.e., historical information sites versus widget shops) necessarily results in very different kinds of sites (form follows function...and marketing follows function too, btw...). Depending upon the category and nature of a site within a category, we might see differences in structure, navigation, content, internal and external linking patterns and a host of very important measures.

Then there are the issues of philosophy, i.e., which kinds of results are better? G has moved in the direction of favoring information over commerce. That makes some of their SERP's cleaner and more useful, but it makes some of their SERP's largely useless, especially when people are searching for products.

Set against these issues, one of the holy grails for any great SE is to be able, regardless of category, to feature the most relevant and useful sites for a given query.

IMO, this is all but impossible for a single algo. So SE's are faced with either hand coding, or segmenting algos in one way or another by category, or finding, if possible, one algo so keen and intuitive that it can show for any given query, excellent results, albeit not perfect ones.

A search engine, working well, should be able to give good if not excellent results across the largest number of categories possible. It should, and hopefully will, someday be able to do this with a minimum of human intervention.

When humans intervene with hand coded SERP's, the problems multiply exponentially. What standards are used? Who makes the choices? Are the choices evaluated by others? If so, who evaluates the choices? How is research done to verify that the choices are good ones? Are the choices regularly updated? Is so how does that process work? Is there a mechanism for the SE to identify worthy sites that were left out of a past hand edit but should probably be included? I'm sure I'm just scratching the surface with these questions.

Hand coding is not consistent, is not reliable, is not subject to sufficient review and vetting, and lacks the brute force credibility that algo's provide when assessing mountains of data without emotion or personal bias.

I'm not arguing against hand coding as a valid way of providing help to Web surfers. I love hand coding, when done in an above-board and consistent way. But when it's done that way, we call it, a directory. ;-)