Penn State University have just released a paper that studies the keyword intent of web searches over a sample of 1.5 million searches.
You can read the entire research paper “Determining the informational, navigational and transactional intent of Web queries” (PDF), but here’s a quick summary:
Their method of classification for each query type was quite complex, and rather interesting to look at - it went a lot deeper than just looking at the search query and making a guess of the intent. As their data source was complete log files (from Dogpile), they were able to track complete search patterns from individual users, so the sequence of searching and the pages ultimately selected by the searcher formed a significant part of the classification process.
I do have a couple of caveats, their ‘informational’ category was essentially ‘what was left’ after they identified the navigational and transactional terms - so that figure is probably inflated somewhat, although not enough to change things significantly I wouldn’t think.
Their ‘Transactional’ category also appears to be a bit broader than I would normally categories it, for instance anything “containing terms related to: movies, songs, lyrics, recipies, images, humor and porn” were considered transactional? personally I would classify nearly all of those as informational (depending on situation - movies and songs could certainly go either way), so from my point of view, 10% of searches being transactional is possibly significantly inflated (hard to tell exactly without seeing the full data).
Another of their observations for informational transactions I thought was quite interesting, they found “a pronounced use of natural language phrases” … something to keep in mind when next trawling through your keyword lists.
If you would like to make a comment, please fill out the form below.
Recent Comments