≡ Menu

Is web performance costing you?

Gawker Media’s radical redesign of all its sites has left a lot of readers unhappy, and they’re apparently leaving in droves. TechCrunch reports that a couple of the sites have suffered major traffic hits, with Gizmodo and Gawker.com losing up to half of their accustomed page views.

The comments from upset readers point to issues with the sites’ new design and architecture, but my hunch is that people will often complain about a website’s design, when what they’re really reacting to is speed and usability. So I ran Webpagetests on Gawker’s three biggest sites, then compared them to archived tests from November 2010.* (Here’s where I thank Pat Meenan profusely for keeping such a great public archive of old tests. It’s really useful to have them around for benchmarking purposes.)

Time to start render: Sites are up to 3X slower

Megablogs typically have abysmal page load times because of ads and third-party content, so let’s focus on start render time, aka the amount of time before usable content appears on the page.

Website Before (November 2010) After (February 2011)
Gawker 2.990 s 5.569 s
Gizmodo 2.067 s 4.142 s
Lifehacker 2.006 s 6.572 s

These sites are two to three times slower now than they were before the redesign. Gizmodo and Gawker have suffered a 50% loss in page views. These numbers roughly correlate to research that finds that:

  • A 1-second delay in page load time equals, on average, 11% fewer page views. (source)
  • Visitors in the top ten percentile of site speed view 50% more pages than visitors in the bottom ten percentile. (source)

If you want a site to feel like an app, it has to be as fast as an app.

Possibly as a way of meeting the opportunities presented by the iPad and similar devices, Gawker seems to want their new websites to perform like apps. This is a bold and cool idea. But if you want a site to feel like an app, it has to be as fast as an app. Users are more likely to assimilate a new interface if it is responsive. New interface + sluggish response times = angry users.

*All tests conducted on IE8/DSL via the server in Dulles,

via webperformancetoday.com

Important stats:
* A 1-second delay in page load time equals, on average, 11% fewer page views.
* Visitors in the top ten percentile of site speed view 50% more pages than visitors in the bottom ten percentile.

Comments on this entry are closed.