Google’s Mouse-Driven Relevancy Patent

Last month Google was awarded a patent on a filing made in 2005. This particular Google patent [#7,756,887] is all about using behavioral data that is used to help refine search results. This month, we’ve been able to corroborate its use in Google one-box (definition) results.

Before we discuss our corroboration of Google’s use of mouse-drive data, let’s be clear about what search results can be influenced by mouse movements and presumably, click-through data. First, go read the patent filing. It’s important to understand that this particular Google patent is:

“A method & system for tuning search result relevancy by observing browser-based activities (mouse hovers over regions & clicks on links). Reordering of results in accordance with respective relevancy values. These relevancy values are used to determine and/or adjust content of a one-box result for search queries.”

Actionable Insight #1:

After reading the patent it’s easy to understand a bit more about Google’s search results tuning processes based on the different mouse activities the search engine monitors, which include, but are not limited to:

  • Observing regions of mouse movement over search result pages.
  • Hover periods; the length of time searchers spend between clicking or not clicking on search result listings.

Google’s search results tuning algorithms allocate different weights to different time periods. For example, hovering over one particular spot of a search results page for a longer period of time is equivalent to searchers forming a more positive opinion about the indexed listing.  So a searcher that hovers over a search engine snippet as if using the mouse to read the listing line-by-line infers a very high level of attention and consequently, more time spent over a particular spot on the search results pages.  (Actionable Insight #1.1: This is also another great reason to write compelling meta descriptions when optimizing your website.)

Actionable Insight #2:

Google doesn’t just use time and mouse behavior to tune its search results.  While the patent clearly indicates that search results are adjusted if hover time periods are greater than threshold values, coefficient values are also aggregated from impression data that details how many times and in what order one-box search results are clicked or not clicked. This is where the flaw in the algorithm can occasionally produce what could be considered suboptimal results.

For example, we recently had a client’s Sitelinks change as a result of surpassing the threshold values for mouse behavior. I know … Sitelinks are not exactly one-box results, but I believe it’s relatively safe to presume that Google Sitelinks are an evolution of one-box results and therefore kept turned by mouse-driven relevancy algorithms, as well are other on-page time-sensitive popularity variants.

The client’s Sitelinks had previously consisted of a series of popular entry pages, bargains and a free catalog. The free catalog Sitelink was replaced by a promotional landing page for a special limited-term offer three days after the content went live on the web … And there it remained in Sitelinks for 10 days after the special offer expired. The content remained listed in Sitelinks, due in part to its mouse-monitored thresholds set by Google.

Having expired event listings in Sitelinks does make for a good user experience, and Google makes itself very clear that its algorithmic search results are all about providing a good user experience. Toward this end, Google makes it relatively easy to correct its algorithmic Sitelink errors.

Google provides webmasters with the ability to remove Sitelink listings via Webmaster Tools, which is exactly what we did for our client in order to have a new Sitelink replace the expired sale page. This is just one of the many ways that Google relies on human intervention to mend the errors of its algorithmic ways.

Make sure your online business has similar access to each of the major search engines webmaster tools in order to be able to take similar fast action – should the need arise.

Actionable Insight #3:

Google’s tuning algorithms for one-box results are also based on coefficient data from “at least one other query-independent metric.” This is just Google’s way of reminding us all that there is more to its “secret sauce” for any particular search result that meets the eye – or in this case, a five-year-old patent filing.

I’d recommend you watch for variants of this patent in other Google search results pages. Who knows where you might see one-box algorithmic thresholds based on mouse-movements in the works? Just look to the left of the page to see what Google’s new Jazz interface if offering you for alternative results or glance to the top and left of the page for advertising positioning. These results listings, in combination with personalized and real-time search, could ultimately be influenced by something as diminutively elusive as a hovering mouse.

Related Resources: