rogerd - 1:50 pm on Oct 24, 2001 (gmt 0)

Great post, Brett.

One other factor I'd add to the clicks vs. position debate is the nebulous area of "quality of results vs. user expectations". Put simply, if the first 5 listings are dead-on for what the searcher is looking for, he may get no farther. If the results contain a lot of irrelevant (from the searcher's viewpoint) stuff, a listing in the teens, twenties, or beyond may get clicked.

This isn't something you can design for, although as Brett notes an on-target description will help you get the clicks you want. Many people are poor searchers, and use terms that generate a result set that has a low density of useful results. If someone searches on "webmaster", who knows what they are looking for? Do they want to hire a webmaster, get a definition of the term, learn about the career, find job openings, get info about a science fiction movie, etc.? Assuming they did have something specific in mind, they may scan through pages of results and only click on a couple of listings. On the other hand, a well-targeted search might produce the same or better information in the first few listings.

I think this variablity is a major reason (in addition to screen geography and other reasons mentioned by Brett) why clicks vs. rank defies accurate statistical analysis, and why tables that claim to predict "clicks per hundred exposures for a given ranking" should carry the warning, "your results may vary - a lot!".