Recap of ‘Personalization, User Data, and Search’ – SES San Jose 2007
MIHMORANDUM NO. 32 | August 27th, 2007
Gord Hotchkiss of Enquiro was the clear ‘rockstar’ on this panel, presenting data from a series of eye-tracking studies his company had completed earlier this year.
His conclusion? Personalized results do appear to work. Time spent on each personalized result, as well as fixation, and clicks were all double when compared to ordinary results.
Interestingly, the embedding of an image or video in the relatively new Universal format causes a phenomenon Gord calls “chunking.” Let’s say for example that an image appears next to a news story in the third slot on the page. People fixate quickly on that third result, and actually eye tracking data shows that results four and five get more views than results one and two. In another session, Gord actually refined his statements about the image-fixation phenomenon. The pictures are actually scanned peripherally, as the real area of focus occurs at the upper-righthand corner of the image, close to the start of the headline. The so-called “Golden Triangle” of visual interest is now re-set below the image chunk, rather than occurring at the top of the page.
Personalization brings in a new twist to the search algorithm: results based on the user, rather than keywords. Personalization works by aggregating cohorts as well as individual web and search history. This means that people in the same region, or people with similar histories, can have their results influenced by others in their cohort. Going forward, search engine optimization will increasingly occur around themes rather than around particular keywords. Dave Davies of Beanstalk went a step further, saying that “trusted” users will affect personalized results within a cohort to a much greater degree than an average user. Group data–membership in online communities, social bookmarking sites, and the like–are also being taken into consideration by some of the other search engines, like Yahoo.
What does this mean for SEOs? Sites need to capture the attention of visitors earlier in the purchasing process, and need to be ‘sticky’ in key users’ web histories and search histories in order to show up at the end of the buying cycle as well. This will seemingly affect spammy made-for-adsense sites, as they’ll be less likely to bring in multiple clicks from the same visitors.
In the Q&A the panelists surmised that Google would not be crawling the same pages multiple times anytime in the near future. This is important to personalized results because it means large sites that are ‘cloaking’ / dynamically serving content depending on IP address are at a disadvantage when they’re up against static pages truly targeted at a particular location.