The tsunami of traffic related to Michael Jackson’s death brought down news web sites on Thursday because those sites had to wait for third-party content, such as images for ads, according to a new analysis.
Keynote Systems said in a report tonight that its analysis of the web’s performance on Thursday showed sites collapsed from the traffic because they were built the wrong way. Those web sites waited for every image to be downloaded from a third-party ad service before displaying the web page to the reader. That led to frustrating delays. Better web sites are built to load a page and then leave a blank spot where it’s waiting for a third-party image to load.
“Our measurement data shows that for sites reported as having performance slowdowns yesterday, internal content delivered quite fast, however content that came from other sources contributed most to the site slowdowns,” said Shawn White, director of external operations at Keynote Systems. “For example, ABC.com’s news content was consistently served in a matter of seconds whereas some of the third party content took much longer.”
White said that, to be prepared for unexpected news events, news sites should require third-party content companies, such as ad networks, to certify the capacity of their networks, perform regular load tests from around the globe, and have strong agreements to get additional server capacity on the fly.
Keynote’s Online News Index includes the following news sites: ABC, AOL, Bloomberg, CBS, CNBC, CNN, CNN Money, Fox Business, Google Finance, Google News, LA Times, MSN Money, MSNBC, MarketWatch, NBC, NPR, NY Times, SF Chronicle, USA Today, Wall Street Journal, Washington Post, Yahoo Finance and Yahoo News. Keynote said that during the peak traffic time, news sites, on average, took nine seconds to load, compared to four seconds before the spike.
The analysis is enlightening. And it means news sites, advertisers, and their technology providers all have to work together to enable fast performance during huge events.
VB's research team is studying web-personalization... Chime in here, and we’ll share the results.