Personalized Search and User Performance Metrics - back in the news
Recently Danny (Sullivan) mentioned Google and the use of performance metrics, which I thought worth discussing.
In his keynote, wonderfully covered by Kalena, he was talking about his vision of Search 4.0 and how personalized search finds its way into the mix. Also of note, the other day Danny was reporting on the keynote interview and Marissa Mayers talk about Previous Query ranking signals coming to the regular index search (not merely AdServing). Personalized search is built upon the concepts that your actions can teach a search engine to better predict your likes and dislikes. Taken on a larger scale, it stands to reason that ranking signals for the regular index can also be mined.
Now what is all this Gobble-D-gook and why do I care? Because it is important.
It means we should once again revisit the world of personalized search and user performance metrics (UPMs). It was last summer when Bill (Slawski) first sent me some very interesting patents relating to UPMs in the regular search index which lead to the whole adventure into personalized search that consumed life for a time last fall. If you didnt catch all the excitement the first time around, its time to listen up.
In many ways looking for ranking signals in the actions performed by the end users is in itself verging on a pseudo-social search in many ways. The more implicit data collection surely falls into behavioural targeting and is a standard practice of qualitative researchers. Of interest, Danny also touched on Social Search which in some ways is what this is
.. more on that next week (social search).
Time for a refresher course
Here is some of the ground covered already to get you up to speed;
Query analysis and user performance metrics;
Probabilistic learning model for text this patent begins to touch on query analysis and probabilistic training of a system that looks for concepts and meanings from the related texts. This is important to get a feel for the query analysis process.
Ranking via user behavioral metrics this starts to look at scoring based on engagement signals and query analysis. This is the one that looks likely as a part of the Previous Query methodology Google is looking to implement. (Coverage from Bill;)
Ad Serving and user performance metrics reading up on this patent is also a good idea as it gives a great deal of insight once more into query analysis and ranking via performance signals.
What every SEO should know about Personalized Search we break down the various elements that make up personalized search from a variety of Google patents.
The Art of personalized Search how to best optimize for personalized search and what webmasters and SEOs need to consider to best leverage it.
Fire Horse Guide to Personalized Search you can download the entire 4 part series in PDF format.
Google and Personalization in Rankings (from Bill)
Keep your head on
One thing that you would be wise to keep in mind is that monsieur Cutts and others have mentioned that such signals are considered generally weak for the regular index. This means one shouldnt start programming a query spam-bot as the likely benefits arent worth the effort. It is in the realm of personalized search that it gets more likely to be a consideration for the SEO community. In the regular index, this would simply lead to a handful of signals out of hundreds.
Do I think it is important to learn? Sure, why else would I be talking here
. if you havent familiarized yourself with query analysis, user performance metrics and personalized search, do so today!
(want to learn more about Personalized Search? Learn from the pros with the SEO Search Engine)