NetMarketShare from Net Applications published a report discussing prerendering of web pages in the Google Chrome browser.

The report stated:

Prerendering in Chrome

The impact of prerendering on desktop usage share

In order to improve performance and usability, browsers are increasingly preloading pages that are never seen by the user.  However, this traffic varies significantly by browser and should not be included in the usage share for the browsers.  Starting in Chrome 13, Google implemented a feature called ‘prerendering’, which loads a hidden page or pages while the user is typing in search queries in order to load that page faster when the user clicks on the associated search result link.  Chrome prerenders pages based on either HTTP headers inserted by the site creator or based on an algorithm that predicts the likelihood the user will click on the search result link.

Chrome has expanded the behavior in version 17 to include search queries typed into the omnibox.

Chrome is the only major desktop browser that currently has this feature, which creates unviewed visits that should not be counted in Chrome’s usage share.  However, the pages that are eventually viewed by the user should be treated normally.

Within the sites in our network, prerendering in February 2012 accounted for 4.3% of Chrome’s daily unique visitors.  These visits will now be excluded from Chrome’s desktop browser share.

Detection Methodology

WebKit’s Page Visibility API

We used a new API in WebKit to detect the visibility state of each page.  With this API, page states can be detected via a modified analytics tracking script.  Pages can start in one of the following states:

State Description
Visible The page is visible to the user.
Prerender The page has been loaded by the browser in an invisible tab and is not visible to the user.
Hidden The page is not visible for a variety of reasons, including being on a tab that does not have focus.

Handling Prerendered Page Transitions

Pages can transition from prerender to visible.  For a prerendered page to count in usage share it must be visible to the user at some point.  The following table details when pages should be counted in usage share:

State/Transition Action
Visible Counts
Prerender (with no transition to visible) Does not count
Prerender to visible Counts

Prerendering state transitions

We only looked at the effect of prerendering on browser usage market share.  But, another consideration is whether as a search user you find prerendering to be beneficial?

For the user the simple benefit of prerendering is page load speed.  Prerendering does require some resources from your processor and bandwidth from your internet connection (probably barely noticeable).  Are the gains in page load speed noticeable?

Is this a reason to use Google search over Bing?

© 2011