Google and Personalized Search Part III : How Personalized Search Works
When last I approached this I was explaining some compelling reasons to give personalized search concepts a second look in your SEO efforts. Most of the reasons related to the increasing pervasiveness of Google Accounts, Personalization and User Performance Metrics. Sure, I know, that sounds all brain-squishy, but its not that hard to get around. To get some perspective read my past posts; Why Google Personalized Search is important and I have seen the future and it is VERY personal .
Let's ring the bell and begin round III shall we?
Google Search is King
Google accounts are pervasive
Google Computers are hot
G-Mobile gives another avenue
Open-Social brings reach
Google TV and Magazines? O come..on...
So Google Battles Microsoft
.for world domination.
And of course
. Google accounts and ultimately, Personalized Search are even more pervasive than ever - Right? Ya follow? Do ya? Good me mateys...moving along.
What is this Personalized Search and Why Should I Care?
Personalization is simply an attempt to deliver more relevant and useful results to the end user (searcher/surfer) and minimize less useful or potentially unlikely results. In a personalized search world, one important element exists that makes it desirable to the end user; the ability to customize, through their actions, or passively, the results that are returned to them in future sessions. Be it of a more useful set of results, or a set of results with less irrelevant, or Spammy entries, it has some attractive features.
Now, let us now begin to imagine how over time this system would also begin to augment the results with related content to selections which you preferred (historically) by trying to understand your surfing, reading and clicking and even scrolling patterns. What now begins to emerge are search engine rankings that are based upon a variety of User Performance Metrics (UPMs) as well as the more traditional ranking mechanisms. A world that is more in the end users control, and less in the hands of marketers or those seeking to manipulate traditional ranking systems.
Maybe someday the Google Toolbar or FireFox Browser (whom Google funds quite well) on your logged in, cross-device/platform Google account, will have something similar to a StumbleUpon thumbs up/down type of application. The below cited patent also describes as; a remove feature selectable object within a web browser application window. So via web interface or ultimately the toolbar, options avail;
Sometimes the search results include a web page that the user deems undesirable. This web page may be deemed undesirable by the user because the web page is spam, the web page relates to content unrelated to the user's interests, the web page contains content that the user dislikes or finds offensive, or for some other reason.
From; Removing Documents
Potentially, the end user may choose to remove an entire site, not merely a single document by; providing a first option for removing the document; providing a second option for removing a site associated with the document.. Is that a resonable assumption? Certainly not. I just want to expand your ability to imagine the possibilities moving forward towards a better understanding.
Regardless of what some search optimizer shaking his fist at the almighty Google overlords may delude themselves into believing, it is not an altogether un-attractive prospect for the searcher on the go. Once again, assuming they realize it is happening at all in their cross device, multiple platform Google Web-world. One does have to consider that a web spammer could only manipulate himself (excuse me sir?) with query spam to rank in such a system where individual users or themed groups of users had some editorial control. It would make a great selling point
added to the previously assumed prevalence from the last time out. Not only do you have the convenience of your multi-platform Google account, but it fights Spam too!! Once again, not un-attractive prospect to the end user or the marketing department.
So what does it mean to the Search Optimizer?
I thought wed look at some of the inner workings and concepts first before devising a battle plan for attack. To begin with, there are a myriad of ways the end users could potentially affect the search results which we can break down to a certain respect at this point. The easiest place to start is by looking at the potential components of a personalized query strategy. In creating system the following metrics could be used in creating personalized ranks;
Previous search queries (probabilistic query refinement); As an example; if the searcher has been recently searching the term diabetes and submits a query for organic food the system attempts to learn and presents additional results relating to organic foods that are helpful in fighting diabetes.
Previously presented results (may be omitted in subsequent queries); results that have been presented to the end user can be omitted in future results for a given period of time in exchange for other potentially viable results.
User query selection (and flagging of similar content); Past selected or preferred documents can be analyzed and similar documents or linking documents can be used to refine subsequent results. Furthermore, certain documents types can be seen as preferred, in what would be a combination of Universal Search concepts. Common websites that accessed can also be tagged as preferred locations for further weighting.
Selection and Bounce rates (and user activity on document/site); an editorial scoring can be devised from the amount of time a user spends on a page, the amount of scrolling activity, what has been printed, or even what has been saved or bookmarked. All can be used to further refine the intent and satisfaction with a given result that has been accessed.
Advertising Activity (performance metrics); the advertisements clicked on can also begin to add to a clearer understanding of the end users preferences and interests.
Some secondary considerations could also involve;
User Preferences (DemoGraphic and Geographic); the end user can also provide specific information as to personal interests or location specific ranking prominence. It could also include favourite types of music or sports, inclusive of geo-graphic preferences such as a favourite sport in a given city.
Historical User Patterns (re-introduction of past removed or further augmentation); a persons surfing habits over a given period of time (eg; 6 months) can also play a role in defining what is more likely to be of interest to them in a given query result. More recent information ( on above factors) is likely to be weighted more than older historical performance metrics within a set of results.
Past visited sites (non SERP activity via account, tool bar or cookie based); many of the above metrics, such as time spent and scrolling on a given web page or historical patterns and preferred locations can also be collected in a variety of ways (invasive or non-invasive). Cookies actually save resources for the Search Engine, an added benefit.
User editorial control (tool based voting system); the end user via toolbar or customized search results, could have the ability to rate search results and sites/pages they visit. This can then be factored into future results in creating probabilistic matches.
Now we have a Blue Print
A few of the key elements surrounding all of this are the historical and probabilistic aspects. By that I mean it is not only sites you have shown an interest in previously, but a virtual educated guessing as to other items that may be of interest based upon the user profile and performance metrics. It is not strictly based upon the end users actions, it seeks to learn. We now have a pretty good idea of what can be involved for delivering the results, we can move onto identifying ways we can account for such issues in the daily grind of optimizing sites and documents for best possible success within these concepts.
I do not see this as a major leap for the SEO enthusiast to deal with. I dare say there are opportunities to leverage personalized search for those that can find the way through these not-so-murky waters. All we need is a little (social) reverse engineering which isnt exactly a foreign term to marketers I am sure.
Thoughts to ponder
Well folks, I can hear the Ship's bell in the Harbour which means duty calls. I will leave this here for you to ponder before my welcome wears thin. A mental note To: Self of late read; make shorter posts --- so we shall stop here. In what I hope to be the last installment on this odyssey into personalization and user performance metrics (UPMs), we will next see what one can do, from an SEO perspective, to best leverage the potential for all this hub-bub in ones ongoing optimization efforts
Until next time.. stay tuned!
Part I; Why Googles Personalized Search is important examining the pervasive nature of Google accounts and Personalized Search now and in the future
Part II; I have seen the future and it is VERY personal Looks at recent business moves by Google and touches on User Performance Metrics
Part IV: the Art of Personalized Search Optimization - a look at ways to leverage it and beyond
More reading; Bill had some interesting additions into the fray recently with; Google and Personalization in Rankings, Google Personalization Methods and Clustering Users for Personalization
Core Patents used during research;
Personalization of WebSearch 20050071328 Mar 01 2005
Variable Personalization of Search Engine Results in a Search Engine 20050216434 - Sept 29 2005
Personalization of placed content ordering in search results 20050240580 - October 27, 2005
Systems and methods for managing multiple user accounts 20060224624 - October 5, 2006
Removing documents 20070043721 - February 22, 2007
Methods and systems for personalized network searching 20050131866 June 16 2005
Systems and methods for providing a graphical display of search activity 20060224938 - October 5, 2006
Systems and methods for providing subscription-based personalization 20060224615 - October 5, 2006
Systems and methods for analyzing a user's web history 20060224583 - October 5, 2006
Systems and methods for combining sets of favourites 20060224608 - October 5, 2006
Systems and methods for modifying search results based on a user's History 20060224587 - October 5, 2006