joined:Dec 26, 2000
I was researching how parallel browsing (tabbed browsing) affects usage statistics and came across this recently published study
[jeffhuang.com] for your information.
Web site and browser design: A number of Websites force the decision of branching on users. Our findings suggest...
...Understanding the Web: We found that the frequency of outclicks and tab switches follow the power law... This can guide design decisions such as the number of hyperlinks to place on a page... Websites that observe deviations from these distributions may want to investigate whether this is caused by spammers or bots...
...Log analysis: ... Studies that consider sessions in search logs as sequential pageviews may be limited because of the prevalence of parallel browsing behavior. Many prior studies in search log mining use sequential timestamp to identify sessions . This ignores the possibility that users may have branched their session, and now have multiple browser tabs which interleaves the click data in server-side logs. Server-side logs used in studies such as , which only capture events visible to the Web server, do not capture the parallel sessions since they do not contain tab data.
...Search interfaces: ...Our finding that informational queries lead to branching more often than navigational queries suggests that a query’s branching percentage may even predict query type. Search engines can apply this by providing different interfaces depending on the user and query...
There's more to it of course. One question I still have is how tabbed browsing affects beacon-generated stats; for example, does a tab switch get look like clicking the back button-- separate discrete views based on the browser focus, or does the session for the hidden page continue as if the user is actually reading it?